default config: rear order-1 streams off larger rivers on a measured width? #327

If done: default decides, on evidence, whether an order-1 stream joining an order ≥ 5 parent rears when its channel width is below the minimum, and on which widths. If never: default keeps declining the exception by inheritance, with no recorded reason, while the bcfishpass config takes it (#324 follow-up).

What bcfishpass does (smnorris/bcfishpass@620dbe4)

• BT, CH, CO, ST and WCT rearing admit an order-1 line with parent order ≥ 5 whose channel width is below rear_channel_width_min (load_habitat_linear_bt.sql:90-97).
• The test sits inside channel_width <= max, so a NULL width is never admitted.

Why that means "measured widths only" in practice

• fwapg models no width for order 1: the cwmodel CTE in fwapg/extras/channel_width/sql/channel_width.sql filters stream_order > 1 (fresh research/fwapg_channel_width.md).
• PARS order 1 by width source: FIELD_MEASURMENT 552, FWA_RIVERS_POLY 4, NULL 19,184, MODELLED 0.
• So bcfishpass's exception admits only order-1 lines with a field-measured (or river-polygon) width. In PCEA, 3,236 of 3,306 order-1 / parent ≥ 5 segments have no width, and none of them rears.
• link does not fill widths (no frs_channel_width() call in R/), so link sees the same NULLs.

The question for default

• Operator's recollection: the modelled width leaves order 1 out because the Thorley & Irvine (2021b) model was not fit on first-order sampling. Find and cite the reason (the channel width memo, attachment 2 of fish_passage_parsnip_2021_reporting).
• Candidate position: admit order-1 / parent ≥ 5 lines on a measured width (FISS or PSCIS), never on a modelled one. That is what bcfishpass does today in effect, made explicit, and it holds if order-1 widths are ever modelled (fresh#234 can now compute them).
• Measure first: km and observation capture on default with and without the exception, on the groups where PCEA: link (bcfishpass config) flags different BT rearing segments than bcfishpass (−8.2 km connection, +30.3 km other) #324 found it moves most (PARS −24.8 km of BT rearing on the bcfishpass config).

Depends on

• fresh's frs_order_child() taking a width test (known width below the minimum; NULL not admitted), which the bcfishpass config needs anyway.

Relates to #324.