Feature: Add --diagnostics option for capturing shell diagnostic logs

[sajeetharan]

Summary

Add a --diagnostics CLI option that writes timestamped diagnostic logs to a file, enabling users to capture shell activity for troubleshooting, performance analysis, and auditing.

Motivation

Currently, CosmosDBShell has limited observability:

• --verbose only controls exception detail level (full stack trace vs. message-only)
• Output redirection (> / >>) captures command output but not execution metadata
• There is no way to trace command execution timing, connection events, or internal errors

Users debugging query performance (RU charges, latency), troubleshooting connectivity, or auditing command history have no built-in mechanism to capture this information.

Proposed Behavior

cosmosdbshell --diagnostics              # log to default path in config directory
cosmosdbshell --diagnostics mylog.log    # log to a custom file path

When enabled, write a structured log file containing:

• Session metadata — timestamp, machine name, OS, runtime version
• Command execution — every command with timestamp
• Command results — OK/FAIL status with elapsed time (milliseconds)
• Errors — exception type and message for failed commands
• Connection events — endpoint URL and connection mode

Sample Output

# CosmosDB Shell Diagnostic Log
# Started: 2026-06-10T14:30:00.0000000-07:00
# Machine: DEVBOX
# OS: Microsoft Windows NT 10.0.22631.0
# Runtime: 9.0.0
--------------------------------------------------------------------------------
[14:30:01.234] [CONNECT ] https://myaccount.documents.azure.com:443/ (mode=Direct)
[14:30:02.456] [CMD     ] dir
[14:30:02.789] [RESULT  ] [OK] 333.2ms | dir
[14:30:05.100] [CMD     ] cat nonexistent
[14:30:05.102] [ERROR   ] cat nonexistent -> ShellException: Item not found
[14:30:05.102] [RESULT  ] [FAIL] 2.1ms | cat nonexistent