Half the r≈2 mass is a wrong scale, not a mislabelled lens, and the fallback draws it at half width

[NewGraphEnvironment]

Problem

inst/notes/terrain-correction.md and the comment on fly_height_ratio_band() say the mass of
film frames beyond r ≈ 2 is a 305 mm lens catalogued as 153 mm, so falling back to nominal
scale is "the correct answer there and not merely the cautious one". Measured while settling
fly#60 by the spacing between frames adjacent by frame number, that is true for about half
of them.

For the 209 near_upper frames beyond r 1.8 catalogued at 153 mm, split per roll-height:

• 90 frames (mostly 1978–81: bc78051, bc79072, bc80048, bc80122, …) give the designed ~60%
  forward overlap under nominal scale and ~80% under the reported height. 14 fit neither. That is the lens
  reading, and the fallback is right.
• 105 of the 209 frames give ~60% under the reported height and ~20% under nominal.
  91 of those are on 1972–76 bc5xxx rolls (bc5701, bc5509, bc5508, bc5699, bc5702, …); 14 are
  not (bc5138 1965, bc79141, bc79043, bc79039, bc78110), and two 1972–76 rolls do not fit this
  way (bc7407 fits nominal, bc7454 neither). Key it per roll, not by year. The height and lens are right and
  scale is the wrong field, recorded at half its true denominator. The fallback draws these
  frames at half their true width, and nothing on the row says so.

The instrument was checked on known answers first: in-band random frames read 0.63
[0.59–0.69], and #54's slipped frames read 0.62 after repair against 0.97 before.

This is the upper-side mirror of fly#60's scale_wrong class, which fly#60 settles for the
lower tail with logbooks as the second witness.

What would settle it

The same two instruments fly#60 used, per roll-height: the logbook height (it should equal the
catalogue's where the scale is the wrong field) and spacing. Rows that pass both go into
flying_height_rolls.csv with factor 1. That needs the table lookup to reach frames above the
band as well as below it; it already keys on the band edge, not on a direction.

Evidence

data-raw/height_calibrate-lower_tail_rolls.R prints the per-roll split ("near_upper r > 1.8
at 153 mm").

Outcome (PR #77)

Settled as proposed, with one change to the rule and two rolls that the spacing split above had wrong:

• The logbook height is not a second witness here. A lens catalogued wrong and a scale catalogued wrong both mean the crew flew the catalogued height, so for this tail (near_upper) the rule also requires spacing to reject nominal scale, not merely fit the reported height.
• 24 roll-heights (120 sampled frames) tabled at factor 1, scale_wrong; 21 (82 frames) log a 12" lens and stay on nominal. Every sampled roll-height is either tabled or listed with its reason in flying_height_rolls_excluded.csv.
• bc79043, listed above among the scale rolls, logs a 12" lens: excluded, witnesses disagree. bc80048, listed among the lens rolls, logs 6" and fits neither reading on spacing: stays on nominal.
• The work read a 600-frame sample, not a census. 1,534 frames on nine 1985 rolls (6,096 m, 305 mm, 1:10000) sit just under the sampled stratum and are likely the same class; they are unmeasured. See inst/notes/terrain-correction.md, "The r ≈ 2 mass is two defects".