Add hardware-accelerated decoding and encoding support

[IanButterworth]

Made possible now that we have JuliaPackaging/Yggdrasil#13288

Developed with Claude

---

Decoding (openvideo / load):

• Add hwaccel kwarg to VideoReader/openvideo for GPU-accelerated decoding (VideoToolbox, CUDA, VAAPI, QSV)
• Frames auto-transferred from HW surface to CPU; identical to SW decoding
• Lazy SwsTransform rebuild when actual HW download format differs from initial guess
• Store swscale/sws_color options on VideoReader so rebuild preserves them
• Add available_hw_devices() public API

Encoding (open_video_out / save):

• Add hwaccel kwarg to VideoWriter/open_video_out/save for HW encoding
• Auto-select HW encoder when hwaccel set and codec_name is nil
• Create hw_device_ctx for codecs needing HW_DEVICE_CTX (NVENC, QSV)
• Add available_hw_encoders() public API

Tests:

• test/hwaccel.jl: decode + encode tests
• util/bench_hwaccel.jl: SW vs HW throughput benchmark

Docs:

• docs/src/reading.md: HW decoding section
• docs/src/writing.md: HW encoding section

---

On my M2 mac

% julia --project util/bench_hwaccel.jl     
Platform HW device: :videotoolbox
HW encoders:        ["h264_videotoolbox", "hevc_videotoolbox", "prores_videotoolbox"]

Resolution  Mode                  Enc (s)   Enc FPS   File MB   Dec (s)   Dec FPS   
------------------------------------------------------------------------------------
640×480     SW encode/decode      0.165     728.1     0.01      0.153     783.4     
            HW enc → SW dec       0.158     759.0     0.02      0.039     3097.6    
            SW enc → HW dec       -         -         -         0.198     604.7     
            HW enc → HW dec       -         -         -         0.231     518.5     

1280×720    SW encode/decode      0.519     231.4     0.02      0.18      666.4     
            HW enc → SW dec       0.404     297.3     0.04      0.108     1109.2    
            SW enc → HW dec       -         -         -         0.409     293.2     
            HW enc → HW dec       -         -         -         0.505     237.6     

1920×1080   SW encode/decode      1.13      106.2     0.02      0.417     287.7     
            HW enc → SW dec       0.851     141.0     0.07      0.241     498.0     
            SW enc → HW dec       -         -         -         0.768     156.2     
            HW enc → HW dec       -         -         -         1.02      117.7 

     Testing Running tests...
┌ Info: Available HW device types (compile-time)
│   hw_devs =
│    1-element Vector{Symbol}:
└     :videotoolbox
┌ Info: Functional HW device types (runtime)
│   working_hw_devs =
│    1-element Vector{Symbol}:
└     :videotoolbox
┌ Info: Available HW encoders
│   hw_encoders =
│    3-element Vector{String}:
│     "h264_videotoolbox"
│     "hevc_videotoolbox"
└     "prores_videotoolbox"
Test Summary:                                                | Pass  Total     Time
VideoIO                                                      | 1239   1239  1m56.4s
  Pointer utilities                                          |   10     10     0.1s
  AVPtr                                                      |   30     30     0.1s
  Reading of various example file formats                    |  140    140    25.9s
  Reading monochrome videos                                  |    5      5     0.7s
  Reading RGB video as monochrome                            |    2      2     2.1s
  IO reading of various example file formats                 |   20     20    34.1s
  Reading video metadata                                     |    9      9     0.2s
  Encoding video across all supported colortypes             |    4      4     0.4s
  Encoding RGBA images (#440)                                |    2      2     0.1s
  Encoding with bare Gray type (#336)                        |    2      2     0.2s
  Missing file extension error (#370)                        |    1      1     0.0s
  Simultaneous encoding and muxing                           |  864    864     2.4s
  Monochrome rescaling                                       |    8      8     0.3s
  Encoding monochrome videos                                 |   12     12     1.1s
  Encoding video with integer frame rates                    |   44     44     0.8s
  Encoding video with rational frame rates                   |    4      4     0.2s
  Encoding video with float frame rates                      |    4      4     0.2s
  Encoding with colorspace and framerate fixes               |   16     16     0.6s
  Video encode/decode accuracy (read, encode, read, compare) |   15     15     5.7s
  Friendly error messages (#360)                             |    1      1     0.0s
  eof correctness for file streams (#320)                    |    4      4     0.1s
  Hardware acceleration                                      |   42     42     2.3s
     Testing VideoIO tests passed 

[codecov]

Codecov Report

❌ Patch coverage is 90.41096% with 14 lines in your changes missing coverage. Please review.
✅ Project coverage is 79.12%. Comparing base (738548b) to head (bf527ff).
⚠️ Report is 2 commits behind head on master.

Files with missing lines	Patch %	Lines
src/encoding.jl	80.76%	10 Missing ⚠️
src/avio.jl	95.74%	4 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #445      +/-   ##
==========================================
+ Coverage   77.87%   79.12%   +1.24%     
==========================================
  Files          10       10              
  Lines        1329     1466     +137     
==========================================
+ Hits         1035     1160     +125     
- Misses        294      306      +12     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.