[bugfix] Find processing instructions on the sibling axes of stored documents

[joewiz]

[This PR was prompted by Joe, drafted by Claude Code, and reviewed by Joe.]

Summary

On a stored document, a processing-instruction test on the preceding-sibling or following-sibling axis never matches anything:

(: stored:  <root><?x 1?><b/><?y 2?><c/><?x 3?></root>  :)
doc("/db/test.xml")/root/c/preceding-sibling::processing-instruction()    (: empty, should be <?x 1?>, <?y 2?> :)
doc("/db/test.xml")/root/b/following-sibling::processing-instruction(x)   (: empty, should be <?x 3?> :)
doc("/db/test.xml")/root/*[preceding-sibling::processing-instruction(y)]  (: empty, should be <c/> :)

preceding-sibling::node() on the same document does return the processing instructions, and constructed documents are unaffected.

Cause

LocationStep.getSiblings handed processing-instruction tests to a VirtualNodeSet, which does not produce siblings for them. A wildcard test on the same axes is matched while streaming the siblings, and both TypeTest and NameTest already match processing instructions on the stream, target name included. So processing-instruction tests now take that path too.

This is the sibling-axis counterpart of 61702d1, which fixed the same VirtualNodeSet hand-off on the self axis.

What Changed

file	change
LocationStep.getSiblings	processing-instruction tests are matched while streaming the siblings, like wildcard tests
sibling-axis-processing-instructions.xqm (new XQSuite)	preceding and following siblings, named and unnamed, in a predicate and around the document element, on a stored document, with the in-memory equivalent

Related issues and PRs

• [bugfix] Look up a named attribute kind test in the attribute index on the self axis #6703 (for [BUG] self::attribute(NAME) is always false on attributes of persistent nodes #6689) fixed the same VirtualNodeSet hand-off for processing-instruction tests on the self axis, in 61702d1. This PR is its counterpart on the sibling axes.
• [BUG] Axis + name test returns wrong result from the 2nd evaluation onwards, when the node was reached through a range-indexed lookup #6690 and its candidate fixes [bugfix] Candidate A for #6690 — reset predicate contexts when a cached index result is reused #6701 and [bugfix] Candidate B for #6690 — keep axis duplicate-suppression state per call #6702 are about named tests on the sibling and preceding/following axes, whose structural-index result is cached across evaluations. This PR changes the branch before that one, which processing-instruction tests take instead. [bugfix] Candidate A for #6690 — reset predicate contexts when a cached index result is reused #6701 also changes getSiblings, but a different branch of it, and the two merge cleanly.
• [bugfix] Apply self-axis kind and wildcard tests in predicates on stored nodes #6769 fixes self-axis kind tests in predicates. Same file, different method; the two merge cleanly.

How this was found

Running the W3C XQuery Update tests against stored documents, which the XQTS runner can now do (eXist-db/exist-xqts-runner#82), showed complex-renames-q1 and complex-replacevalues-q3 failing only on stored documents. Neither failure was in the update: the verification query's preceding-sibling::processing-instruction() found nothing even before any update. The context is summarized on #6214. BaseX 12.4 gives the expected result for each case above against a stored database.

Test Plan

• sibling-axis-processing-instructions.xqm: all pass with the fix; the stored-document tests fail without it
• xquery.CoreTests
• full exist-core test suite
• qt3tests HEAD, with all five of these stored-document fixes applied together, against current develop: no regressions (27,613 tests compared; the only three changes are tests that also vary between develop runs)

🤖 Generated with Claude Code

[github-actions]

📊 XQTS result comparison

Comparison of this run against develop.

Metric	develop	this run	Change
🟢 Passed	28,864 (90.72%)	28,866 (90.73%)	+2 (+0.01 pp)
🟢 Failures	1,550	1,548	-2
➖ Errors	134	134	0
➖ Skipped	1,267	1,267	0
🧪 Total tests	31,815	31,815	0

Relative to develop: 3 newly passing, 1 newly failing, 0 new errors, 0 newly skipped — counting only tests recorded in both runs whose outcome changed.

🔴 Newly failing tests (1)

• Constr-inscope-2 (was passing)

🟢 Newly passing tests (3)

• Constr-inscope-3 (was failing)
• Constr-inscope-1 (was failing)
• Constr-inscope-4 (was failing)

_{Runtime: 410.6s (-39.38s vs develop).}