Improve SensitivityMSM robustness and test coverage

[aman-coder03]

Proposed changes

follow ups on the review comments from #925
This addresses several non-blocking follow-up suggestions from the review of the SensitivityMSM implementation.
changes focus on improving robustness, numerical stability, and test coverage without changing the public API or the underlying MSM estimation logic

Specifically,

• replaces the unsupported-learner blocklist with an explicit allowlist of supported learner families (S-, T-, and DR-learners), so unknown learner types fail safely instead of being treated as supported
• expands test_SensitivityMSM to cover S-, T-, and DR-learners instead of only the T-learner
• makes the MSM tests deterministic by seeding the synthetic data generation using RANDOM_SEED
• clips propensity bounds before division in the MSM bound computation to avoid numerical issues when propensity estimates are extremely close to 0 or 1

The documentation wording regarding the theoretical sharpness of the bound and the minor double-fit optimization are intentionally left for future work, as discussed in the previous review

Types of changes

What types of changes does your code introduce to CausalML?
Put an x in the boxes that apply

• Bugfix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation Update (if none of the other choices apply)

Checklist

Put an x in the boxes that apply. You can also fill these out after creating the PR. If you're unsure about any of them, don't hesitate to ask. We're here to help! This is simply a reminder of what we are going to look for before merging your code.

• I have read the CONTRIBUTING doc
• I have signed the CLA
• Lint and unit tests pass locally with my changes
• I have added tests that prove my fix is effective or that my feature works
• I have added necessary documentation (if appropriate)
• Any dependent changes have been merged and published in downstream modules

Further comments

If this is a relatively large or complex change, kick off the discussion by explaining why you chose the solution you did and what alternatives you considered, etc. This PR template is adopted from appium.

[jeongyoonlee]

The allowlist matches type(learner).__name__ against base-class names, so concrete S/T/DR subclasses are rejected: XGBTRegressor, MLPTRegressor, LRSRegressor, and XGBDRRegressor all hit the NotImplementedError here despite being supported. Use isinstance(learner, (BaseSLearner, BaseTLearner, BaseDRLearner)) instead of name membership — it covers subclasses and still rejects X/R and unknown types.

The test_SensitivityMSM parametrization uses only the base classes (all in the allowlist), so this isn't caught — please add a concrete subclass (e.g. XGBTRegressor) to the parametrization.

[jeongyoonlee]

LGTM