feat(sampling): allow dynamic tuning steps (Fixes #8080)

[Jayant-kernel]

This PR allows step methods to declare their required number of tuning steps, instead of relying on the hardcoded default of 1000. This is efficient for samplers like [CategoricalGibbsMetropolis]

(cci:2://file:///c:/Users/jayan/Desktop/pymc/pymc/step_methods/metropolis.py:687:0-865:38) which require 0 tuning steps.
Changes:

• Added default_tune_steps attribute to [BlockedStep]
  (cci:2://file:///c:/Users/jayan/Desktop/pymc/pymc/step_methods/compound.py:103:0-238:56) (defaults to None -> 1000).

• Set default_tune_steps = 0 for [CategoricalGibbsMetropolis]
  (cci:2://file:///c:/Users/jayan/Desktop/pymc/pymc/step_methods/metropolis.py:687:0-865:38) and BinaryGibbsMetropolis.

• Updated pm.sample to resolve tune=None by checking the step methods.
  I verified this with new unit tests and by confirming that CategoryGibbsMetropolis now produces traces with 0 warmup steps by default. @ricardoV94 this is ready for your review.

Related Issue:

markdown

• Closes Allow steps to decide number of tune #8080
  Checklist:

markdown

• Checked that the pre-commit linting/style checks pass
• Included tests that prove the fix is effective or that the new feature works
• Added necessary documentation (docstrings and example notebooks)

[Jayant-kernel]

@ricardoV94 @aloctavodia
Thanks! I've opened a PR for this here: #8110.

The implementation allows steps to declare their own tuning needs (e.g., Gibbs tune=0) and falls back to the default otherwise.
Ready for your thoughts whenever you have a moment.