Add OpenOffice.org 1.x / StarOffice XML (.sxw/.sxc/.sxi/.sxd) read+write support

[Mearman]

Add read+write support for the OpenOffice.org 1.x / StarOffice XML formats: .sxw (Writer), .sxc (Calc), .sxi (Impress), .sxd (Draw), and their .stw/.stc/.sti/.std template counterparts.

Unlike the other legacy formats surveyed alongside this one (WordPerfect #819, Lotus SmartSuite #820, Microsoft Works #818), this format genuinely does have a public specification: "OpenOffice.org XML File Format 1.0", published by OpenOffice.org/Sun and the direct document OASIS based ODF 1.0 on. It's XML-in-a-ZIP, the same shape odf.js already parses for the ODF successor formats (.odt/.ods/.odp/.odg) — an older, pre-OASIS schema variant of the same lineage, not a new binary format needing its own parser from scratch. The bulk of the work is mapping that older schema's namespaces and element/attribute differences onto what odf.js already reads, closer to a variant reader than a new hand-written codec.

Part of the same legacy-format survey as #85 (doc/xls/ppt), #83 (RTF), #818 (Works), #819 (WordPerfect), and #820 (Lotus SmartSuite).

[Mearman]

Read half is done in #839 (draft): readSxw/readSxc/readSxi/readSxd in odf.js, built the way this issue suggested — a package transform into the ODF shape, then the existing readOdt/readOds/readOdp/readOdg. Validated against five genuine 2003-2005 documents pulled off the Wayback Machine, not just the spec.

The framing here holds up, with one correction: it's a bit more than namespace renaming. The real structural change is that OpenOffice.org 1.x gives every style a single style:properties where ODF has typed style:text-properties/style:paragraph-properties/style:table-cell-properties/… — and you can't just copy the attributes into all of them, because fo:background-color means a character highlight, paragraph shading or a cell fill depending on which element it sits in. That needs a real per-family split (LibreOffice's own algorithm: ordered candidate list, first match wins). Bare draw:image/draw:text-box needing the draw:frame wrapper ODF introduced is the other one. Everything else genuinely is renames plus a few value respellings ("inch" → "in" is load-bearing — parseOdfLength rejects inch outright).

The write half needs its own issue, or this one narrowing. It's not blocked on anything OpenOffice.org-specific: odf.js has no ODF content writer either (its README: "every read-side construct has no write-side counterpart because this package's typed layer is read-only"), so an inverse transform has nothing to target. Writing .sxw is really "build an ODF writer, then invert the transform" — much bigger than this reader, and it belongs to ODF first.

One thing worth recording for anyone touching this later: OpenOffice.org's own retained DTD (xmloff/dtd/nmspace.mod) declares the presentation: namespace as http://openoffice.org/2001/presentation. That's wrong. Every real .sxi declares .../2000/presentation, and the 2001 spelling appears nowhere in LibreOffice's namespace table, so nothing could ever have read it. Trusting the DTD there silently loses every presentation attribute.

[Mearman]

odf.js now has a genuine .sxw writer (writeSxw/writeSxwContent, PR #878) built on top of the ODF-to-OOo1 inverse transform and the existing writeOdt/writeOdtContent pair. Verified round trip against the existing, unmodified readSxw reader.

.sxc/.sxi/.sxd remain read-only, since odf.js has no writeOds/writeOdp/writeOdg yet for the inverse transform to build on -- so this doesn't close the issue, just covers the .sxw quarter of it.

[Mearman]

odf.js now has a real .ods writer (writeOdsContent/writeOds, PR #889), the direct prerequisite for .sxc write support the same way the .odt writer was the prerequisite for .sxw's (PR #878). .sxc's own OOo1x inverse-transform write path hasn't been built yet -- that's the next step to actually close the .sxc quarter of this issue.

[Mearman]

odf.js now has a real .sxc writer too (writeSxc/writeSxcContent, PR #894), built on the .ods writer + the same ODF-to-OOo1 inverse transform .sxw uses -- the transform turned out already generic across every ODF media type, so no changes to it were needed. Two of the four formats (.sxw, .sxc) now have full read+write. .sxi/.sxd remain read-only, blocked on odf.js having no writeOdp/writeOdg yet.

[Mearman]

Started on the odp/odg write gap: a real writeOdpContent/writeOdp writer for odf.js's typed layer (the direct prerequisite for .sxi write support, the same way writeOds was for .sxc) is in progress now. writeOdg will follow the same pattern once that lands, then the two thin .sxi/.sxd wrappers on top of both, mirroring exactly how .sxw/.sxc each landed as writer-then-wrapper.