Add profiling/tracing module and speed improvments to correspondence function of optimizer

[akenmorris]

• Add basic profiling and tracing infrastructure
• Speed improvements for correspondence function of optimizer

This PR adds a very basic profile and trace capability to ShapeWorks. It is controlled with environment variables SW_TIME_PROFILE=1 and SW_TIME_TRACE=1. When profiling is on, a file, profile.txt is written at the end of execution and displayed on the screen as well. It looks roughly like this:

THREAD 0:
------------------------------------------------------------------------------------------------------
   %Time    Exclusive    Inclusive   #Calls   #Child    Exclusive    Inclusive Name
             total ms     total ms                        ms/call      ms/call
------------------------------------------------------------------------------------------------------
   100.0           96      2785689        1        4           96      2785689 shapeworks
    96.4         1046      2684521        1        1         1046      2684521 Optimize::RunOptimize
    96.3            4      2683476        1     5000            4      2683476 GradientDescentOptimizer
    96.3         5283      2683472     5000    10000            1          537 optimizer_iteration
    91.3           21      2542943     5000     5000            0          509 gradient_before_iteration
    91.3       743136      2542922     5000    25000          149          509 CorrespondenceFunction::ComputeUpdates
    49.0      1363987      1363987     5000        0          273          273 correspondence::covariance_multiply
    10.4       290396       290396     5000        0           58           58 correspondence::gramMat
     4.9       135245       135245     5000        0           27           27 parallel_sampling
     3.5        97351        97351        1        0        97351        97351 Optimize::SetParameters
     3.4        95425        95425     5000        0           19           19 correspondence::lhs_rhs
     1.4        40078        40078     5000        0            8            8 correspondence::svd
     0.4         9899         9899     5000        0            2            2 correspondence::pinvMat
     0.1         2216         2216        1        0         2216         2216 set_up_optimize
     0.1         1504         1504        1        0         1504         1504 Optimize::Initialize
------------------------------------------------------------------------------------------------------

All times are in milliseconds.

Exclusive total time : Time spent in this timer and not any child timers/calls
Inclusive total time : Time spent in this timer and also any child timers/calls
#Calls : Number of times this timer/function is called
#Child : Number of child/sub-calls to other times/functions
Exclusive ms/call : Mean exclusive time per call
Inclusive ms/call : Mean inclusive time per call

The API is quite simple:

    TIME_START("correspondence::gramMat");
     ...
    TIME_STOP("correspondence::gramMat");

Or:

{
TIME_SCOPE("CorrespondenceFunction::ComputeUpdates");
...
}

TIME_SCOPE uses RAII to stop the timer at the end of the scope.

Tracing produces a trace.json file at the end. This uses the google trace event format (https://docs.google.com/document/d/1CvAClvFfyA5R-PhYUmn5OOQtYMH4h6I0nSsKchNAySU/preview?tab=t.0#heading=h.yr4qxyxotyw) and can be viewed in tools such as Perfetto UI (https://ui.perfetto.dev/)

Additionally, profiling of the correspondence function was performed and numerous improvements were made, primarily by converting VNL code to Eigen:

Before:

--------------------------------------------------------------------------------------
   %Time    Exclusive    Inclusive    #Call   #Subrs    Inclusive  Name
                 msec   total msec                        ms/call
--------------------------------------------------------------------------------------
   100.0           95      5115882        1        4      5115882 shapeworks
    98.0         1025      5014928        1        1      5014928 Optimize::RunOptimize
    98.0            4      5013903        1     5000      5013903 GradientDescentOptimizer
    98.0         6766      5013899     5000    10000         1003 optimizer_iteration
    95.2           22      4871999     5000     5000          974 gradient_before_iteration
    95.2      1495283      4871977     5000    10000          974 CorrespondenceFunction::ComputeUpdates
    39.9      2040669      2040669     5000        0          408 correspondence::gramMat
    26.1      1336025      1336025     5000        0          267 correspondence::covariance_multiply
     2.6       135134       135134     5000        0           27 parallel_sampling
     1.9        97173        97173        1        0        97173 Optimize::SetParameters
     0.0         2215         2215        1        0         2215 set_up_optimize
     0.0         1471         1471        1        0         1471 Optimize::Initialize
--------------------------------------------------------------------------------------

After:

------------------------------------------------------------------------------------------------------
   %Time    Exclusive    Inclusive   #Calls   #Child    Exclusive    Inclusive Name
             total ms     total ms                        ms/call      ms/call
------------------------------------------------------------------------------------------------------
   100.0           96      2785689        1        4           96      2785689 shapeworks
    96.4         1046      2684521        1        1         1046      2684521 Optimize::RunOptimize
    96.3            4      2683476        1     5000            4      2683476 GradientDescentOptimizer
    96.3         5283      2683472     5000    10000            1          537 optimizer_iteration
    91.3           21      2542943     5000     5000            0          509 gradient_before_iteration
    91.3       743136      2542922     5000    25000          149          509 CorrespondenceFunction::ComputeUpdates
    49.0      1363987      1363987     5000        0          273          273 correspondence::covariance_multiply
    10.4       290396       290396     5000        0           58           58 correspondence::gramMat
     4.9       135245       135245     5000        0           27           27 parallel_sampling
     3.5        97351        97351        1        0        97351        97351 Optimize::SetParameters
     3.4        95425        95425     5000        0           19           19 correspondence::lhs_rhs
     1.4        40078        40078     5000        0            8            8 correspondence::svd
     0.4         9899         9899     5000        0            2            2 correspondence::pinvMat
     0.1         2216         2216        1        0         2216         2216 set_up_optimize
     0.1         1504         1504        1        0         1504         1504 Optimize::Initialize
------------------------------------------------------------------------------------------------------

This reduced the runtime by 46% (1 hour 25 minutes down to 46 minutes).

This was a particular fixed domains runtime with 2048 particles + normals, about 150-ish shapes. Similar speedups will not be seen at lower particle/shape counts and no improvement occurs for initialization/sampling.