fix: add DOI to prefix_mapping in _get_not_found_ids

[lior-airis]

Summary

• _get_not_found_ids was missing DOI from its prefix_mapping dict, causing false "IDs not found" warnings for every DOI-prefixed lookup (e.g., DOI:10.1145/792550.792552)
• The bare DOI value was added to found_ids without the DOI: prefix, so it never matched the prefixed input ID
• Added 'DOI': 'DOI' to the mapping
• Changed matching logic to always add bare external ID values alongside prefixed forms, so both DOI:10.1145/... and 10.1145/... inputs match correctly (preserving backward compatibility with bare DOI inputs used in existing tests)
• Added two unit tests (sync + async)

Reproduction

from semanticscholar import SemanticScholar
sch = SemanticScholar()
# This returns the paper correctly, but also logs a false warning:
# WARNING:semanticscholar:IDs not found: ['DOI:10.1145/792550.792552']
papers = sch.get_papers(['DOI:10.1145/792550.792552'])

The paper data IS returned, but the spurious warning is confusing and the return_not_found=True parameter falsely includes DOI-prefixed IDs in the not-found list.

Test plan

• Added test_get_papers_doi_prefix_not_false_positive (sync)
• Added test_get_papers_doi_prefix_not_false_positive_async (async)
• Existing tests pass (bare DOI inputs like 10.2139/ssrn.2250500 still match correctly)

[lior-airis]

Closing in favor of #116 which fixes the Python 3.14 nest_asyncio incompatibility first. Will rebase the DOI prefix fix on top of that and reopen as a separate PR.