Name the posted bit in terminate()

[mgrossmann]

The first of the two leftovers noted at the end of #112. The second turned out
not to exist — see below.

The change

terminate() tested the ECB posted bit as a raw 0x40000000U, twice, while the
wait loop in main() a few hundred lines up spells it ECB_POSTED_BIT. Same
bit, same struct field, two vocabularies — and the raw one is the harder to grep
for when the question is "where do we check whether the socket thread has ended".

Generated assembler is byte identical before and after (cc370 -O1 -S, cmp);
clibecb.h defines the macro as exactly that constant.

make clean, make test-host 78/78, tools/check-module-data.py clean.

The other leftover was a false alarm

#112 also said "the bare-__asm__ pattern fixed here exists in other FTPD
sources — this PR only covers src/ftpd.c". That is wrong, and worth correcting
rather than leaving as a standing to-do.

src/ftpd.c is the only file in the repository containing inline assembler
at all — five STIMER sites, all of which already carry
"0", "1", "14", "15" from #114:

src/ftpd.c:227  src/ftpd.c:296  src/ftpd.c:477  src/ftpd.c:622  src/ftpd.c:777

Nothing else in src/, include/ or asm/ has an __asm__ beyond the
SETC '&FUNC' name markers. So there is no follow-up owed here.

The equivalent sweep in libc370, where the pattern really does recur, is
mvslovers/libc370#134.