New transcoding logic implementation

[tumb1er]

• [x] 1. Split source file to small chunks to lower task run time
• [x] 2. Upload chunks to temporary distributed storage
• [ ] 3. Run a set of tasks to transcode all chunks with same profile
• [x] 4. Concatenate transcoded chunks to a resulting file and segment to HLS
• [ ] 5. Upload result to storage
• [x] 6. Transcode audio stream in one pass without splitting, merge with video chunks
• [ ] 7. Fix metadata damaged while analyzing HLS chunks and manifests

[tumb1er]

Does not work yet: audio obtains silence gaps while splitting. New strategy:

• don't split audio at all, transcode it while downloading and splitting source
• merge audio and multibitrate video together while concatinating video chunks.

[tumb1er]

Also: instead of synchronizing chunk tasks, implemented resumable transcoding: chunks are not transcoded again after task restart.