Task: Draw the ER diagrams as SVG

The diagrams ship as vector files and no environment-specific raster or image map stays in git.

• The draw step emits SVG; ores_schema.png and ores_schema.cmapx are deleted and untracked.
• The CMake target, the script and the Java arguments produce the vector output.
• The conventions doc and the memory note stop telling a reader to leave the PNG and cmapx alone.

(Implementation strategy. Written when work starts; key decisions are distilled into the parent story's * Decisions at close, but the plan itself stays — it is the historical record of what we did.)

Every number this task claims about a tree, with the command that produced it and the commit it was measured at. A number without both is not a measurement, because a clean-up record describes a tree and trunk moves under a branch: a consumer count, a gate's component count and a suite's size all change with no code changing, and the reader of a bare number cannot tell which tree it came from. Where the raw output matters, put it in an evidence/ file beside the task and name that file in the Command column.

Manual QA scenarios (scaffolded via compass add test_scenario) that verify this task. Link new ones here as they're created; the scenario doc itself links back via its "Verifies task" field.

The drawing step emits SVG and every generated diagram is drawn, not only the index: the script expands ores_schema*.puml and the CMake target does the same through bash -c, because CMake cannot expand a glob in a COMMAND and the file set has to follow the model rather than a list kept in the build description. 22 diagrams draw in one pass, 4.5 MB in total, and a pass writes no PNG and no cmapx.

ores_schema.png and ores_schema.cmapx are deleted from the index. The SVGs that replace them are committed: they are what the documentation site renders, through proj: links that export as image tags pointing at the file in the repository, so an image outside the tree is a broken page. The first cut of this task ignored them, which would have published a schema page with no schema on it.

ORES_JAVA_ARGS is shared with the sixteen component diagrams that still draw PNGs, so it keeps PLANTUML_LIMIT_SIZE; the ER target passes it through and adds -tsvg. The script's own invocation drops the limit, since a vector has no pixel count to overflow, and its java fallback does the same.

The conventions doc names the drawn files, links the index and every group image so the site shows them, and the memory note now counts the SVGs among the addresses a model change has to regenerate.