test: add SIGTERM delay tests for shell, Java, and Go processes

[vlsi]

Verify that the entrypoint's SIGTERM_EXIT_DELAY mechanism works correctly by asserting console output.

Each child process prints a "delayed message" after 5 seconds;

• with delay=0 the message is absent (child killed before printing),
• with delay=8 it is present (child alive during the delay window).

[lis0x90]

@vlsi I'm curious, why not just extend existing test for signal propagation https://github.com/Netcracker/qubership-core-base-images/tree/main/tests/signal-propagation?

[vlsi]

1. The existing test deliberately avoids SIGTERM (

   qubership-core-base-images/tests/signal-propagation/sample-process.sh

   Line 3 in 576b439

   # use SIGUSR1 to avoid sleeping 10 seconds in entrypoint trap handler for SIGTERM

   ). The new test is specifically about exercising that avoided path.
2. Different structure — sigterm-delay runs two containers per process type (delay=0 vs delay=8) and tests three child process types (shell, Java, Go), making it significantly heavier (~30s vs ~5s).
3. Different concern — signal-propagation checks that signals reach the child and exit codes propagate. sigterm-delay checks that the delay mechanism keeps the child alive during the grace period.

Do you still think it is worth merging the tests?

Frankly, I am leaning towards keeping two independent tests as they would be easier to review/understand/maintain/drop in the future.