Split tokenizer state-machine responsibilities

[dv00d00]

Summary

• split ordinary tag-state handling from tag token lifecycle code
• isolate the fused tag-tail scanner in its own partial file
• separate tokenizer configuration, write-loop, and output helpers
• preserve the tokenizer instruction stream; this is a structural-only refactor

Validation

• whitespace-insensitive reconstructed token streams match the original Tags and Common files
• dotnet build AngleSharp.ReadOnlyDom.slnx -c Release --no-restore -m:1 (0 warnings)
• 179,562 / 179,562 net10.0 tests passed via the TUnit executable
• UTF-8 token smoke: 57 documents / 38,831,169 bytes, 0 failures
• UTF-8 DOM projection and QQ scraper checks passed
• CSharpier check passed for all six touched files

The repository-wide CSharpier check still reports pre-existing drift in unrelated files tracked separately by #57.