fix: deliver the fork delta since the 1110 sync wave - #865

What this changes

Delivers the fork delta accumulated since the #859 delivery as one batch commit on top of
master (the single-batch delivery the fork sync policy prescribes). Two lines, delivered
whole:

• The start page and library stay usable at their edges
  (fix: keep the start page and library usable at their edges (audit #562) axisrow/BeamCraft#564, from the 2026-11-09 visual audit Visual bug audit 2026-11-09: 55 defects confirmed on screen, sorted by priority axisrow/BeamCraft#562, items
  23–25 and 54–56): long author/environment names stay on one line with an ellipsis instead
  of pushing the way back off the start window; an unreadable settings file shows a short
  refusal word in the title bar with the reason on hover; at the declared minimum size the
  selected-model panel scrolls instead of clipping its State row and Open model button; the
  welcome combo truncates a long environment selection; stale and wrong texts are fixed (the
  folder note that followed the user to other tabs, an empty All models claiming a folder was
  chosen, a refusal kept after its fields changed, the Blank hint contradicting
  docs/design/blank-environment.md); and library cosmetics (template cards share one
  height, row texts cut with an ellipsis instead of a mid-glyph clip, the preview retry is a
  themed button outside the permanent triangle budget, the CREATED column right-aligns and
  gives way on a narrow table).
• The BOM and schedule reports are honest about cuts, zeros and precision
  (fix: make the BOM and schedule reports honest about cuts, zeros and precision axisrow/BeamCraft#567, audit items 3, 39 and 40, on their report surfaces only): a straight
  cut member's report length is the extent of its re-evaluated certified mesh projected onto
  the axis, so a beam cut from 4000 to 2000 mm no longer prints the authored length and a
  non-physical derived area; a known non-zero value that rounds below the unit's display
  precision keeps the noise-stripped full value instead of printing 0; profile and mark
  labels sort naturally (PL8 before PL20); and schedule-export renders every standard
  quantity column at the BOM's display precision, so one quantity has one text across the
  reports (extends the [P2] beamcraft-cli: dashed and dash-dot circles export as solid lines; S still reads as 5; schedule and BOM round mass differently axisrow/BeamCraft#490 mass parity).

The fork-only policy surface (its workflow forms, docs/plan/, visual evidence, fork tooling)
is not part of this delivery; the upstream-owned Python CI tools are untouched. Bare fork
issue references in the touched comments and docs are qualified as axisrow/BeamCraft#N
(including three #490 mentions an earlier delivery left bare).

How you verified it

• cargo fmt --all -- --check (recorded gate stage, exit 0)
• cargo clippy --workspace --all-targets -- -D warnings (recorded gate stage, exit 0)
• cargo test --workspace (recorded gate: 2724/2724 runnable tests executed with nextest
  0.9.149 on macOS aarch64, plus mandatory doctests and cargo doc -D warnings, all exit 0)

Never-crash

Every model file, IFC/DSTV import and keystroke is untrusted input (AGENTS.md §4).

• No new unwrap/expect/panic!/todo! outside tests
• No indexing or slicing with input-derived positions; allocations sized by input are capped
• No unsafe

Provenance

Read AGENTS.md §1 and §2 before adding any asset.

• Nothing taken from Tekla Structures or any Trimble product (assets, code, documentation, tables)
• No GPL/AGPL code copied in
• No new assets

Anything a reviewer should look at first

• The report-length change applies only to straight members: polybeams and curved members
  keep their authored path length, and the derived section area is computed from the same
  re-evaluated mesh extent, so the non-physical 30000 mm2 reading cannot return.
• The zero display keeps real zeros as 0; only a non-zero value that rounds below the
  unit's display precision falls back to the full value.
• The natural label order feeds both the element table and the summary groups of one export,
  so the two tables cannot disagree again; text comparisons in the schedule filter go through
  the same natural order.

🤖 Generated with Claude Code