Subaligner

* Support Python 3.9 and TensorFlow 2.5
* Add the subaligner_batch CLI for batch alignment on nested folders
* Tidy up dependencies and update docs

* Support subtitle translation both during and post alignment
* Add the `--translate` flag for specifying the language code pair
* Add the `--languages` flag for displaying language codes used for translation and stretch
* Handle input file paths containing whitespace
* Update dependencies for future Python 3.9 support

* Support more subtitle formats, including Scenarist Closed Captions (.scc), Youtube SubViewer (.sbv) and Transcript (.ytt) formats
* Improve logs on the error stack trace and logging levels
* Add the custom build for Docker Hub and update installation instruction

* Allow passing in `sound_effect_start_marker` and `sound_effect_end_marker` for ignoring sound effects during training
* Tie each docker build to the then release tag
* Cover extraction on frame rates with decimal places
* Fix bugs on shifting subtitles in TTML and auto-detection of Unicode encodings
* Require passing in the frame rate on converting to subtitles in MicroDVD
* Improve container image builds

* Support more subtitle formats, including SAMI (.sami, .smi) and EBU STL (.stl)
* Add the docker builds for Fedora and Debian
* Allow alignment against embedded subtitles using selectors
- `embedded:page_num` for DVB Teletext
- `embedded:stream_index` for Matroska
* Add subtitle file encoding auto-detection
* Fix XML declarations for TTML files
* Add the `subaligner_convert` CLI for subtitle conversion
* Make the ffmpeg/libav path re-configurable via env var `FFMPEG_PATH`

* Support alignment on remote media files
* Allow language configuration for subtitle stretching
* Add type hints for developers
* Support Tensorflow 2.4.0
* Fix and Update docs
* update docker images and deprecate Ubuntu 19.*