fix: guard set_scale against invalid concurrency_modifier returns

[Swigler]

Summary

• Validate concurrency_modifier return value in set_scale() — reject None, non-int, and values < 1 (defaults to 1 with a warning log)
• Wrap the callback in try/except — a raising modifier keeps the current concurrency instead of crashing the entire worker

Fixes #458

What was wrong

If a user-provided concurrency_modifier callback returns None (or any non-integer), self.current_concurrency is set to None. This later crashes at the job scheduling loop:

TypeError: '<' not supported between instances of 'int' and 'NoneType'
  File "rp_scale.py", line 241, in run_jobs
    while len(tasks) < self.current_concurrency and not self.jobs_queue.empty():

The crash kills the worker, causing 2-3 minutes of downtime while the Docker container restarts.

What changed

rp_scale.py — set_scale() method:

• Wrap concurrency_modifier() call in try/except — if it raises, log a warning and keep the current value
• Validate the return: must be int and >= 1, otherwise default to 1 with a warning

test_scale.py — 5 new tests:

• None return → defaults to 1
• Negative return → defaults to 1
• Zero return → defaults to 1
• Exception in modifier → keeps current concurrency
• Valid int → applied correctly

Test plan

• All 9 tests in test_scale.py pass (4 existing + 5 new)
• All 17 tests across both modified test files pass

[Swigler]

Ping — if concurrency_modifier returns an invalid value, set_scale crashes the worker. This adds a guard. Ready for review.