Add integration tests for process_html_file with flite

[Tomotz]

Add integration tests for process_html_file with flite

Summary

Adds 9 end-to-end integration tests for process_html_file that actually invoke the flite binary. Tests are intentionally loose to allow changes in output format/reductions without breaking:

• No exceptions: basic HTML, empty paragraphs, HTML entities, stdout mode all run without error
• IPA present: output contains IPA characters
• Output resemblance: original English words appear in output (word overlap ratio > 50%)
• Structure preserved: HTML tags, non-paragraph content, multiple paragraphs handled correctly

All tests are gated behind a @requires_flite skip marker — they are skipped automatically if the flite binary isn't built.

Review & Testing Checklist for Human

• CI will silently skip all 9 new tests because the workflow doesn't checkout submodules or build flite. I couldn't push workflow changes (OAuth scope limitation). To make these tests run in CI, manually update .github/workflows/tests.yml to add submodules: true to the checkout step and a cd flite && ./configure && make -j$(nproc) build step.
• test_multiple_paragraphs asserts >= 6 <p> tags for 3 input paragraphs (assumes each paragraph produces an IPA + original pair). Verify this matches the intended process_html_file contract and is loose enough for future format changes.
• Run git submodule update --init && cd flite && ./configure && make -j$(nproc) then python -m pytest test_main.py::TestProcessHtmlFileIntegration -v locally to confirm all 9 tests pass on your machine.

Notes

• Link to Devin run
• Requested by @Tomotz