`PCHIPInterpolation` for 2 data points is broken

[SouthEndMusic]

I understand that for a degree > 1 interpolation having only 2 data points doesn't make sense, but I feel like this edge case should be dealt with better than a vague index error:

julia> PCHIPInterpolation([1.0, 1.0], [1.0, 2.0])
ERROR: BoundsError: attempt to access 1-element Vector{Float64} at index [2]
Stacktrace:
  [1] throw_boundserror(A::Vector{Float64}, I::Tuple{Int64})
    @ Base .\essentials.jl:14
  [2] getindex
    @ .\essentials.jl:916 [inlined]
  [3] (::DataInterpolations.var"#_du#55"{…})(k::Int64)
    @ DataInterpolations C:\ProgramData\DEVDrives\.julia\packages\DataInterpolations\TCMJ8\src\interpolation_utils.jl:284
  [4] _broadcast_getindex_evalf
    @ .\broadcast.jl:678 [inlined]
  [5] _broadcast_getindex
    @ .\broadcast.jl:651 [inlined]
  [6] getindex
    @ .\broadcast.jl:610 [inlined]
  [7] macro expansion
    @ .\broadcast.jl:973 [inlined]
  [8] macro expansion
    @ .\simdloop.jl:77 [inlined]
  [9] copyto!
    @ .\broadcast.jl:972 [inlined]
 [10] copyto!
    @ .\broadcast.jl:925 [inlined]
 [11] copy
    @ .\broadcast.jl:897 [inlined]
 [12] materialize
    @ .\broadcast.jl:872 [inlined]
 [13] du_PCHIP(u::Vector{Float64}, t::Vector{Float64})
    @ DataInterpolations C:\ProgramData\DEVDrives\.julia\packages\DataInterpolations\TCMJ8\src\interpolation_utils.jl:314
 [14] #PCHIPInterpolation#32
    @ C:\ProgramData\DEVDrives\.julia\packages\DataInterpolations\TCMJ8\src\interpolation_caches.jl:1343 [inlined]
 [15] PCHIPInterpolation(u::Vector{Float64}, t::Vector{Float64})
    @ DataInterpolations C:\ProgramData\DEVDrives\.julia\packages\DataInterpolations\TCMJ8\src\interpolation_caches.jl:1341
 [16] top-level scope
    @ REPL[14]:1
Some type information was truncated. Use `show(err)` to see complete types.

[dmetivie]

I am facing exactly the same issue with different interpolation methods. In my case
The error is with RegularizationSmooth also very vague
ERROR: ArgumentError: invalid Array dimensions
In some other case with the same method I was getting
ERROR: BoundsError: attempt to access 2-element Vector{Float64} at index [3]

I think in the first case I was feeding something like u = [missing, missing, missing, 1.0] while in the second case it was u = [1.0, 20.].
(This method use cubic interpolation, so at least 4 (or 3?) points are necessary.)

I think all methods needs to check before doing anything else the number of valid data points, to throw an appropriate error in this is not respected.

We could define an internal function like

min_number_of_points(m::InterpolationMethod) = 

for each method and then test

length(filter(!ismissing, u)) >= min_number_of_points(m)

The property min_number_of_points could be displayed in the docstring and the table proposed in Issue #368