Add audio decoding benchmarks

[NicolasHug]

Example output:

~/dev/torchcodec (add-audio_benchmark*) » python benchmarks/decoders/benchmark_audio_decoders.py --path test/resources/nasa_13013.mp4.audio.mp3            
Benchmarking nasa_13013.mp4.audio.mp3, duration: 0:00:13, averaging over 30 runs:
torchcodec.AudioDecoder                  med = 4.49, mean = 4.76 +- 0.78, min = 4.28, max = 8.39 - in ms
torchaudio.load(backend='ffmpeg')        med = 5.01, mean = 5.15 +- 0.79, min = 4.51, max = 8.98 - in ms
torchaudio.load(backend='sox')           med = 7.21, mean = 7.54 +- 0.70, min = 7.00, max = 10.62 - in ms
torchaudio.StreamReader                  med = 7.34, mean = 7.76 +- 0.70, min = 7.23, max = 10.01 - in ms

[scotts]

I ❤️ benchmarks!