Make the shared RLE sequence limit configurable

[Shubham-Padkonde]

The RLE sequence limit is currently repeated as 50 in normalization, translation, and VCF output. Add a shared GA4GH_VRS_RLE_SEQ_LIMIT environment setting so these paths agree, as requested in #598.

The default remains 50 bases. A non-negative integer selects the limit, 0 omits optional sequence values, and none disables the limit. The setting is read at import time; invalid values produce an error naming the setting. Explicit normalization/translation arguments still override the default. Documentation includes a CLI example.

Validation on Python 3.12/Linux:

• Eight new subprocess tests exercise actual normalization, translation, and VCF output with recorded SeqRepo responses, plus invalid settings. Five fail before the change; all pass afterward.
• Full suite: 145 passed with the bundled SeqRepo fixture configured and VCR recording disabled.
• Ruff lint/format and repository pre-commit checks pass.

Prepared with Codex assistance. Fixes #598.

Running the three affected modules alone produces three existing VCR cassette failures; the same failures reproduce against unchanged source. The full suite above passes.

[korikuzma]

@Shubham-Padkonde Thanks for making this PR! I'd like to see the tests changed before approval/merge.

[korikuzma]

I think it'd be good to have docs here too

Suggested change

	"""Read the maximum optional RLE sequence length from the environment."""
	"""Read the maximum optional RLE sequence length from the environment.
	Set `GA4GH_VRS_RLE_SEQ_LIMIT` environment variable to change
	the maximum length of optional `ReferenceLengthExpression.sequence` values.
	The default is `50` bases. Use a non-negative integer, `0` to omit the optional
	sequence, or `none` (case-insensitive) for no limit.
	:raise ValueError: If provided value is not an integer or is a negative value
	"""

[korikuzma]

I'm wondering if we could avoid using subprocess with a full script (it's a little hard to read and I imagine it could be hard to maintain). Could we instead look into mocking the environment variables in our existing tests (such as

vrs-python/tests/extras/test_allele_translator.py

Line 755 in 859e77a

def test_rle_seq_limit(tlr):

and https://github.com/ga4gh/vrs-python/blob/main/tests/extras/test_annotate_vcf.py)?

[jsstevenson]

yeah generally the recommended way to injectj/manage env vars in tests is with the pytest monkeypatch module

https://docs.pytest.org/en/stable/how-to/monkeypatch.html#monkeypatching-environment-variables