fix: the six minor findings of the YAML study review

[0rkhann]

Fixes the six minor findings left from the final review of #28. Each has a test that failed before the fix.

Finding	Now	Test
Missing study file, product: [P] or studied_species: [S] crash with a traceback	Reported as an invalid study, exit 1	test_missing_file_is_a_study_error, test_product_and_studied_species_must_be_names
The studied species can be the product	Error	test_studied_species_cannot_be_the_product
Every Ea fit skipped gives KeyError: 'c0(S)'	Clear message (sample earlier), nothing empty cached, exit 2	test_flux_sign_handling
check misses warnings that run gives	check prints negative-barrier warnings and ΔG for typed energies; typed energies no step uses are a warning	test_check_shows_typed_energy_warnings, test_unused_typed_energy_is_a_warning
A species name starting with "via" is misread	Parsed correctly	test_species_name_starting_with_via
A re-run reusing the barrier tables prints no ΔG or warnings	Same messages as the first run	test_rerun_reports_the_same_messages

docs/USAGE.md documents the new message and warnings. All 13 test files pass; ruff 0.16.10 is clean; both examples pass check.

🤖 Generated with Claude Code