fix: read XLSX tags independent of namespace prefix

[Aias00]

Purpose of the pull request

Closed: #1156

Valid OOXML namespace prefixes are arbitrary, but the XLSX SAX handlers previously dispatched only unprefixed, x:, and ns2: tag names.

What's changed?

• Enable namespace-aware SAX parsing.
• Normalize worksheet and shared-string tags through localName, with a qualified-name fallback.
• Add a compatibility test that rewrites worksheet and shared-string XML with a p: prefix and verifies model reading.

Verification:

CompatibilityTest#readXlsxWithArbitraryNamespacePrefix
Tests run: 1, Failures: 0, Errors: 0, Skipped: 0
spotless:check: passed

Checklist

• I have read the Contributor Guide.
• I have written the necessary doc or comment.
• I have added the necessary unit tests and all cases have passed.

[nkuprins]

Dispatching on localName alone matches elements from other namespaces. Before, only unprefixed names and the x: and ns2: prefixes matched.

POI's XSSFSheetXMLHandler checks the namespace first:

if (uri != null && !uri.equals(NS_SPREADSHEETML)) {
    return;
}

Should we add the same guard, plus a test?

[Aias00]

Done in cfde3ec. Both worksheet and shared-string SAX handlers now ignore elements outside the SpreadsheetML namespace. The integration fixture also includes foreign-namespace row/c/v and si/t elements, so both paths are covered. CompatibilityTest passes all 10 tests, and spotless:check passes.

[skytin1004]

Hi @Aias00, while testing nested namespace content, I noticed that the namespace guard skips the foreign element events, but characters() still receives their text while the outer <t> is active. I tested an outer <t> containing a foreign <ext:t> element, and Name + foreign + 0 was read as Nameforeign0. I saw the same issue inside <v>, where the extra text can change the shared-string index and cause an IndexOutOfBoundsException.

I think both handlers should keep an ignored-element depth: increment it when entering a foreign element, ignore all character data while the depth is nonzero, and decrement it on the matching end event. Adding nested foreign-content cases for both <t> and <v> would cover this path.

[Aias00]

Done in d9ea4c2. Both SAX handlers now track ignored-subtree depth, so character data inside a foreign namespace is discarded until the matching end element. The integration fixture covers nested foreign <ext:t> inside SpreadsheetML <t> and nested <ext:v> inside <v>, in addition to sibling foreign elements. The targeted regression test and all 10 CompatibilityTest cases pass, and spotless:check passes.