Subaligner

* Add `subaligner_train` and `subaligner_tune` CLIs for advanced users
* Allow advanced users to pass in their trained models during subtitle synchronisation
* Update documentation on installation with pipx
* Fully support TF2 Optimizers on training
* Fix keyboard interrupt within the training process
* Add the instruction for Windows users on running subaligner as a container

* Support more subtitle formats, including (Advanced) SubStation Alpha (.ssa, .ass), MicroDVD (.sub), MPL2 (.txt) and TMP subtitles (.tmp)
* Add the docker build for ArchLinux
* Add the option of building an app as a pex binary
* Support TensorFlow 2.3
* Add the `fail_on_segment` flag
* Support running subaligner as a script with `python -m`
* Bug fixings

* Save intermediate files into the temp folder
* Fix the log level configuration
* Add `--quiet` and `--output` flags and integration tests on command-line interface
* Skip frame rate extraction on alignment with audio input
* Improve locking mechanism and speed up frame rate extraction
* Improve memory efficiency and throttle concurrent job submissions

* Support docker builds for el7, el8, u18, u19 and u20
* Support Pipenv
* Add profiling
* Upgrade dependencies and bug fixing
* Add docs for contribution

* Advance to Python 3.7 and Python 3.8
* Support Tensorflow 2.2.0
* Release to PyPI