hal: Gate the loadusr "Waiting for component" pacifier on a tty

[grandixximo]

Fixes #4287.

Problem

The pyvcp test (and any test comparing captured output) flakes on slow or busy CI:

+Waiting for component 'mypanel' to become ready....
 mypanel float OUT 0 mypanel.dial-a-out
 ...

The "Waiting for component..." pacifier is printed by halcmd's loadusr -Wn to stderr after 2 s of waiting. It lands in the test's result file anyway because Ubuntu 24.04's xvfb-run executes the wrapped command with "$@" 2>&1, merging stderr into stdout (Debian's and upstream xorg's xvfb-run keep the streams separate). On a loaded ubuntu CI runner, pyvcp takes >2 s to become ready, the pacifier fires, and the merged output breaks the compare against expected. The sed '/mypanel/!d' in the test keeps the line because it contains the component name.

I reproduced this byte-identically on a Debian host by extracting Ubuntu noble's xvfb-run (xvfb_21.1.12-1ubuntu1.8_amd64.deb), delaying pyvcp startup with a sleep wrapper, and saturating the CPU.

Fix

The pacifier is a progress indication for an interactive user; in batch use it is pure noise. Only print it when stderr is a terminal, in both halcmd and halrmt. The warn_newly_loaded_comps diagnostic (wrong -Wn name hint) is unaffected and still prints.

Also fixes halcmd printing the terminating newline from count >= 100 even though the pacifier itself only starts at count == 200.

Testing

• Ubuntu xvfb-run + delayed pyvcp + saturated CPU: was XFAIL with the exact CI diff, now passes with no output pollution.
• Pacifier still appears when stderr is a tty.
• Full runtests suite: 291/291 pass.