What does this PR do? Please describe:
This PR extends the CLI with support for tab completions. Initially, I tried using argcomplete, but it was slow and ran a Fairseq2 process each time I pressed tab, which was far from ideal.

I then discovered shtab, a package that generates shell tab completion scripts efficiently. According to its README:

• What: Automatically generate shell tab completion scripts for Python CLI apps
• Why: Speed & correctness. Alternatives like argcomplete and pyzshcomplete are slow and have side-effects
• How: shtab processes an argparse.ArgumentParser object to generate a tab completion script for your shell

After implementing shtab, tab completion works beautifully. Here are some examples:

$ fairseq2 [TAB]
assets        llama         lm            mt            wav2vec2_asr  

$ fairseq2 ll[TAB]
$ fairseq2 llama

It’s especially useful for things like presets. Instead of listing presets and copying them manually, you can now use tab completion:

$ fairseq2 lm [TAB]
chatbot               generate              instruction_finetune  preference_finetune   
$ fairseq2 lm c[TAB]
$ fairseq2 lm chatbot

It also works with arguments:

$ fairseq2 lm chatbot --[TAB]
--cluster               --help                  --model                 --temperature           --top-p                 
--dtype                 --max-gen-len           --seed                  --tensor-parallel-size  

Tab completions are specific to each shell and require the installation of corresponding scripts in the shell's directory.

Initially, I intended to extend the project by adding post-install scripts to automatically set up tab completion scripts (b63af98). However, I discovered that this approach only worked with zip and tarball distributions, while fairseq2 uses wheel builds. As a result, I reverted that change.

Currently, you can enable tab completions manually by using --print-completions {bash|zsh|tcsh}. The output of this command can be saved to the shell's completion scripts directory for persistence or evaluated directly for the current session.

For example:

eval "$(fairseq2 --print-completion bash)"

Please note: each time the CLI interfaces are modified, the shell completion scripts need to be updated.

Does your PR introduce any breaking changes? If yes, please list them:
None.

Check list:

• Was the content of this PR discussed and approved via a GitHub issue? (no need for typos or documentation improvements)
• Did you read the contributor guideline?
• Did you make sure that your PR does only one thing instead of bundling different changes together?
• Did you make sure to update the documentation with your changes? (if necessary)
• Did you write any new necessary tests?
• Did you verify new and existing tests pass locally with your changes?
• Did you update the CHANGELOG? (no need for typos, documentation, or minor internal changes)