Make server e2e readiness timeout configurable

[quartzjer]

Summary

• add PARAKEET_SERVER_E2E_READY_TIMEOUT to tests/server_e2e.sh
• validate the timeout value before launching the server
• set the CI server-e2e readiness timeout to 600s while keeping the job-level 20-minute timeout as the outer cap
• include the configured timeout in the failure message

Why

PR #41 hit a server-e2e failure where the server was still downloading the first-run tdt_ctc-110m-q4_k.gguf model at the previous 180s readiness deadline. The log showed the 125 MB download around 82% when server_e2e: server not ready fired, then finishing seconds later. Since /health only becomes available after model download and load, the readiness wait needs to allow hosted-runner network variance.

Validation

• bash -n tests/server_e2e.sh
• tests/server_e2e.sh; test $? -eq 77
• PARAKEET_SERVER_E2E=1 PARAKEET_SERVER_E2E_READY_TIMEOUT=bogus tests/server_e2e.sh /bin/true; test $? -eq 1
• PARAKEET_SERVER_E2E=1 PARAKEET_SERVER_E2E_READY_TIMEOUT=0 tests/server_e2e.sh /bin/true; test $? -eq 1
• git diff --check

[mudler]

looking good, thanks!