fix: preserve unexpected peak refinement failures

[RoboX2020]

Addresses the findpeaks.py part of #115.

Polynomial refinement now skips only ValueError and NumPy LinAlgError failures. It emits a RuntimeWarning, pointing to the caller, instead of printing to stdout. Unrelated exceptions and user interrupts are no longer swallowed.

Added a real boundary-neighbourhood regression and injected-error tests for RuntimeError and KeyboardInterrupt. All three assertions fail with the original bare except.

Validation on Python 3.11:

• Peak suite: 52 passed plus 2 subtests, on OpenCV 4.12.0.88 and 5.0.0.93.
• Full suite on OpenCV 5: 961 passed, 97 skipped, 7 subtests passed.
• Both changed files pass black; tests pass ruff. Two existing source lint findings (unused eps and a wildcard import in the demo) remain unchanged.

Full suite on OpenCV 4 was not run. Other exception sites in #115 are left to separate changes.

[codacy-production]

Up to standards ✅

🟢 Issues 0 issues

Results:
0 new issues

View in Codacy

🟢 Metrics 0 complexity · 0 duplication

Metric	Results
Complexity	0
Duplication	0

View in Codacy

_{NEW Get contextual insights on your PRs based on Codacy's metrics, along with PR and Jira context, without leaving GitHub. Enable AI reviewer
TIP This summary will be updated as you push new changes.}

[petercorke]

Thanks — turning the swallowed fit failure into a RuntimeWarning and letting real errors through is just right. Merged.