fix: Replace exec liveness probe with tcpSocket for LMCache cache server

[Asthenia0412]

Description

The LMCache cache server livenessProbe uses an exec probe that runs /workspace/LMCache/examples/kubernetes/health_probe.py, but this script does not exist in the LMCache v0.5.4 Docker image. The cache server communicates via TCP socket protocol, not HTTP.

Changes

• Changed the liveness probe from exec to tcpSocket in deployment-cache-server.yaml
• Updated the corresponding Helm test in cacheserver_test.yaml

A tcpSocket probe is more appropriate because:

1. The LMCache cache server uses a TCP socket protocol
2. No external script is required - Kubernetes can check the port directly
3. Works with all LMCache versions without depending on a specific script

Closes #1057

[gemini-code-assist]

Code Review

This pull request replaces the Python-based exec liveness probe in the cache server deployment with a simpler tcpSocket probe, updating the corresponding test cases as well. The reviewer recommends using a named port (e.g., 'caserver-cport') instead of referencing the port directly in both the deployment template and the test assertions to align with Kubernetes best practices.

[gemini-code-assist]

Using a named port (e.g., caserver-cport) instead of referencing the port number directly in the tcpSocket probe is a Kubernetes best practice. This decouples the probe configuration from the specific port number, making the chart more maintainable and robust if the port number changes in the future.

            tcpSocket:
              port: caserver-cport

[gemini-code-assist]

Update the test assertion to expect the named port caserver-cport instead of the hardcoded port number 8000 to align with the template change.

          tcpSocket:
            port: caserver-cport