BREAKING(xml): expose doctype and processing instructions from parse

[tomas-zijdemans]

parse() validated processing instructions and the DOCTYPE, then threw both away. This PR keeps them: the doctype lands on XmlDocument.doctype, and PIs become tree nodes. This addresses the gap noted at the end of #7212.

XML 1.0 §2.6 is blunt about PIs: they "MUST be passed through to the application". Comments outside the root element were dropped the same way, so a license header above <root> vanished on round-trip. Also fixed.

What changed

• New XmlDoctype type, exposed as XmlDocument.doctype. Captures the name, public id, and system id. Only produced with disallowDoctype: false. The secure default still rejects DOCTYPE outright.
• New XmlProcessingInstructionNode in the XmlNode union, plus an isProcessingInstruction guard. PIs inside the root land in children.
• PIs and comments outside the root land in new prolog/epilog fields on XmlDocument, in document order, omitted when empty.
• ignoreProcessingInstructions moved from ParseStreamOptions to BaseParseOptions so the sync parser accepts it too. Streaming behavior is unchanged.
• stringify() emits all of the above and gains a doctype: boolean option (default true), mirroring declaration.

Breaking changes

• XmlNode gains a member: exhaustive switches over node types need a processing_instruction case.
• PIs now appear in children by default. Opt out with ignoreProcessingInstructions: true.
• A DOCTYPE after the root element, or a second DOCTYPE, now throws XmlSyntaxError (XML 1.0 §2.8). Both were silently accepted before, which becomes observable once the doctype is stored. Only reachable with disallowDoctype: false.

Fidelity notes

• The DTD internal subset is validated but never captured (DTD content is deliberately unprocessed), so parse → stringify drops it.
• Doctype/PI interleaving is not preserved: a PI written before <!DOCTYPE> re-serializes after it. Still valid per §2.8.
• <!DOCTYPE foo PUBLIC "pub"> (public id, no system id) is leniently parsed and re-serialized as-is, matching the DOM XMLSerializer.

Testing

46 new tests across parser, serializer, and type guards, including round-trips: XHTML doctype, xml-stylesheet PI, internal-subset drop. One existing test asserted the old discarding behavior and was rewritten. deno task ok passes.

The two commits split cleanly: doctype first, PIs second, in case one half needs to land alone.

Built with assistance from Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 97.41379% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 95.03%. Comparing base (cdb83c6) to head (27efa94).

Files with missing lines	Patch %	Lines
xml/stringify.ts	95.52%	2 Missing and 1 partial ⚠️

Additional details and impacted files

@@           Coverage Diff            @@
##             main    #7261    +/-   ##
========================================
  Coverage   95.03%   95.03%            
========================================
  Files         618      618            
  Lines       51596    51954   +358     
  Branches     9340     9434    +94     
========================================
+ Hits        49035    49376   +341     
- Misses       2021     2032    +11     
- Partials      540      546     +6     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.