Dynamic PTS for test runs

[sansmoraxz]

• Remove hard coded path which can break running cli apps (as these apaths are used by terminals). Use tmp based paths instead.
• Ensure these paths are cleaned up after test run (regardless of state)
• glide cli is not really necessary with model go. So that was cleaned up too
• Moved this test into seperate script file

Fixes: #128

[alexjoedt]

pkill -P $$ kills only direct children of the current shell. If diagslave or socat spawns child processes, those will be orphaned. Consider using a process group kill (kill -- -$$) or tracking PIDs explicitly.
The || true suppresses all errors, including unexpected ones, which may hide issues in CI.

[sansmoraxz]

They don't spawn any but still good point.

[alexjoedt]

sleep 0.5 is a race condition
The sleep 0.5 waiting for socat to create the PTY symlinks is a classic timing hack. On a loaded CI machine this can be too short. A more robust approach would be to poll until the symlink appears, e.g. until [ -e "$tmpdir/pty0" ]; do sleep 0.05; done with a timeout guard.
Also in line 30

[alexjoedt]

rtu and ascii do the same thing except for the -m flag. We could avoid the copy-paste by using one shared branch/helper and just changing the MODE passed to diagslave.

[alexjoedt]

MODE and TEST_FILTER are used without much validation. MODE already falls through to the usage error if it’s missing, but an empty TEST_FILTER becomes -run "", which runs all tests. If that’s intended, maybe we should document it.

[alexjoedt]

set -e won’t catch failures from background processes. If diagslave or socat fails to start, the later go test may fail with a less clear error. It would be easier to debug if we check that the background process is actually running before continuing. Wdyt?