fix: decode percent-encoded unreserved chars before resolving dot segments in normalize

[gaoflow]

Summary

URL.normalize() resolves dot segments (. and ..) before decoding percent-encoded unreserved characters. Per RFC 3986 Section 6.2.2.2, this order is incorrect -- unreserved characters should be decoded first.

For example, %2e is the percent-encoded form of . (a period is an unreserved character). The path /%2e%2e/foo should be normalized to /foo, but currently it normalizes to /../foo.

Before

>>> URL.from_text("http://example.com/%2e%2e/foo").normalize().to_text()
'http://example.com/../foo'

After

>>> URL.from_text("http://example.com/%2e%2e/foo").normalize().to_text()
'http://example.com/foo'

Test plan

• Added test_normalize_percent_encoded_dot_segments covering three cases:
  • %2e%2e (double-dot parent reference)
  • %2e (single-dot current directory reference)
  • foo/%2e%2e/bar (encoded parent traversal in middle of path)
• All 102 tests pass (101 existing + 1 new)

This PR was written under human direction (gaoflow).