docs(source_coverage): config_tu/ is out of scope on purpose, not by omission

[andrewboudreau]

Comment only, no behaviour change. Filed because the comment was actively
misleading in a gate, which is the failure class this tree keeps getting bitten by.

What it said vs. what is true

module_of() carried:

# Any other layout (config_tu/, a future module root): name it by its directory, so
# an unrecognised path is still attributed rather than silently merged with arm9.

But config_tu/ can never be passed to module_of(). Both scanners are hard-scoped to
CONFIG_DIR = "config":

entries_from_tree()  ->  (root / CONFIG_DIR).rglob("delinks.txt")
entries_from_ref()   ->  git ls-tree -r <ref> -- CONFIG_DIR

So the function advertised handling for a root neither caller can reach.

Why that is worth a commit

The obvious "fix" on reading it is to widen the scanners so the config_tu/ branch
finally does something. That would be wrong, and tu_config.py says so in its own
contract:

This generator writes a PARALLEL root (config_tu/ by default) [...] nothing here can
reach the ROM build: config/**/delinks.txt remains the sole authority over what
mwccarm compiles

rombuild.py reaches that root only behind an opt-in --module flag, for the TU-shaped
treemap experiment.

I measured what widening would actually buy, rather than arguing it:

config_tu/**/delinks.txt on main
delinks files	8
entries	29
entries marked complete	0

Zero. complete is the whole ballgame — with it mwccarm compiles the source, without it
dsd supplies the object from ROM bytes. So widening the scan finds no source-built
byte at all; it imports 29 unenrolled entries and the false alarms that come with them.
The new comment records that number so the next reader does not have to re-derive it.

Why I was looking at this file

Chasing the same question that produced #2046: what actually notices when dsd stops
compiling a range from source and hands it back to the retail cartridge. config_tu/
came up as a candidate blind spot. It is not one — it is correctly out of scope — and
this commit is the difference between "out of scope" and "looks like an oversight."

Verification

tools.test_source_coverage        Ran 25 tests   OK
python tools/source_coverage.py   2,067,148 B across 74 module(s), 10,973 range(s)
                                  .init 125,712 B   .text 1,941,436 B   exit 0

Byte-for-byte the figure main reports, which is the point: a comment change must move
nothing. Pre-push green: port_refcheck 405/405, check_src_tu 90/90,
duplicate-sources 11199 stems none doubled. No compiled source touched, so there is no
byte claim here and none is made.

[tangos-validator]

✅ PR validation — Passed

noverify: no source/build-data changes in this PR

_{Each changed src/*.c|*.cpp is compiled and its relocated bytes compared to the binary data on a private build box. Passing requires every changed file to reproduce the ROM byte-for-byte with correct relocation targets — this catches WRONG-DEST relocations and non-reproducing near-misses that ledger-scoped linkcheck skips.}