fix(view): say N other beads over what rooting the forest moved

[GraemeF]

The collapsed line that stands over everything Shift+F takes off the screen said 3 roots held back. Three faults in six words: it named what had been done to those roots rather than where they now are, it implied a reason the program cannot know, and it spelled them with a word readers do not use. It now says 3 other beads, and 1 other bead where there is one.

Nothing on the line says why they are behind it. An earlier version of this change carried a second clause explaining that, and it is withdrawn rather than reworded: why the reader asked for one bead is theirs, and a line guessing at it would be a claim bdi cannot support.

The number still counts the roots behind the line while the word says beads, and that is deliberate. A root whose tracker refused holds no beads at all, so counting beads would draw 0 beads over a root that is really there and really unreadable. Each thing counted is a bead in its own right, and the neighbouring cells keep the arithmetic honest without arithmetic: 3 other beads · 2 agents beneath · 1 bead beneath cannot be read as three beads of which one wants looking at, because the one that wants looking at is said to be beneath them.

Nothing is coined, so docs/design.md's terminology table gains no entry. The line is the shape phrase::elided already draws.

GroupKind::OutOfTheWay and the layout walk under it are named for what the group is rather than for what the row says, so a future rewording of the sentence moves no identifiers.

The same row carried a defect this change also fixes. Its cells abutted — 2 agents beneath⚠ 1 bead beneathF for the whole forest on the shipped screen — because this was the one state block on a row not built through draw::beside, whose own doc comment claims every one of them is. It rides here rather than in a bead of its own because it is three calls in the function this change was already rewriting. Both existing tests asserted on each cell alone, which passes whether the cells abut or not, so the assertion that holds the gap now reads the joined row.

The fix costs four columns and the new wording gives them back, so the row is no wider than it was. The block gains two gaps at two columns each, and 3 other beads is four columns shorter than 3 roots held back, as 1 other bead is than 1 root held back. The key hint F for the whole forest therefore still survives from 77 columns up, which is where it survives from on main.