ci: Release automation

[ko3n1g]

What does this PR do ?

Implements a manual workflow_dispatch that allows us to build, test, and publish a PyPi wheel as well as create GH release (tag).

It is secured by a GH environment that runs on main (protected) only by a PyPi token that is resolved in this environment only. Put different: If one tries to use this token on a branch other than main it will be empty. So to make use of this token, they need to be able to merge their script into main, which of course only works by review.

The release page will be populated by the CHANGELOG.md file (awesome that you already have that!!). I extract the section from (incl) to (excl) ## ... blocks as the description of the release page.

Changelog

• Please update the CHANGELOG.md under next version with high level changes in this PR.

Usage

• You can potentially add a usage example below

# Add a code snippet demonstrating how to use this 

Before your PR is "Ready for review"

Pre checks:

• Make sure you read and followed Contributor guidelines
• Did you write any new necessary tests?
• Did you add or update any necessary documentation? Make sure to also update the NeMo Framework User Guide which contains the tutorials

Checklist when contributing a new algorithm

• Does the trainer resume and restore model state all states?
• Does the trainer support all parallelism techniques(PP, TP, DP)?
• Does the trainer support max_steps=-1 and validation?
• Does the trainer only call APIs defined in alignable_interface.py?
• Does the trainer have proper logging?

Additional Information

• Related to # (issue)

[ko3n1g]

@terrykong could you have a second look?

[ko3n1g]

Create a protected environment to hold our PyPi secrets: https://github.com/NVIDIA/NeMo-Aligner/settings/environments

[ko3n1g]

Merging this now since I incorporated @terrykong 's feedback and also tested the shell commands locally during the last release. It ultimately can only be tested "live" anyway, so there's no point in holding this PR back