Guard wildcard similarity for large bodies

[maurosoria]

Summary

• Add a conservative content-size guard before the ambiguous wildcard similarity fallback.
• Skip the expensive body similarity calculation when either response body is larger than 256 KiB.
• Add regression tests proving very large bodies return before content_similarity is called while medium-size bodies still use the fallback.

Why

The ambiguous wildcard fallback is useful for dynamic soft-404 pages, but full body similarity can become unnecessarily expensive on very large responses. For large bodies, this PR keeps the heuristic conservative and avoids the costly comparison without disabling the fallback for ordinary and moderately large HTML error pages.

Validation

• python -m unittest tests.core.test_scanner
• python -m unittest tests.core.test_scanner tests.core.test_wordlist_backend tests.core.test_dictionary_templates
• python dirsearch.py -u https://example.com -w tests/static/wordlist.txt -q
• git diff --check