fix: return non-zero exit code when collection errors occur

[hughhan1]

Summary

• The run_tests() function now returns exit code 1 when collection errors occur
• Previously, errors were displayed but execution continued if some tests were collected
• This matches the behavior of the main CLI entry point (main_cli_with_args)

Test plan

• Added integration test for collection errors with enum subclass scenario
• All existing tests pass (102 Rust tests, 48 Python collection integration tests)

[github-actions]

Benchmark Comparison Report

Comparing current results against baseline from main branch.

Repository	Benchmark	Baseline	Current	Change	Status
click	CLI startup time	36.9ms	37.4ms	+1.3%	[-]
click	Test discovery performance	45.3ms	45.4ms	+0.1%	[-]
flask	CLI startup time	35.4ms	35.1ms	-1.0%	[-]
flask	Test discovery performance	43.3ms	45.1ms	+4.2%	[-]
flask	Test execution with pytest runner	1.72s	1.75s	+1.4%	[-]
more-itertools	CLI startup time	31.7ms	31.9ms	+0.4%	[-]
more-itertools	Test discovery performance	43.2ms	44.0ms	+1.9%	[-]
more-itertools	Test execution with native runner	3.28s	3.40s	+3.7%	[-]
more-itertools	Test execution with pytest runner	3.90s	3.04s	-21.9%	[++]
pydantic	CLI startup time	28.6ms	33.5ms	+17.0%	[!!]
pydantic	Test discovery performance	127.2ms	141.9ms	+11.6%	[!!]

Summary

• Total benchmarks: 11
• Regressions (>5.0% slower): 2
• Improvements (>5.0% faster): 1
• No significant change: 8

Performance regressions detected!
Please review the results above to ensure these are expected.

Legend

• [++] Significant improvement (>10.0% faster)
• [+] Minor improvement (5.0-10.0% faster)
• [-] No significant change (±5.0%)
• [!] Minor regression (5.0-10.0% slower)
• [!!] Significant regression (>10.0% slower)