fix(bqjdbc): optimize formatter in BigQueryJdbcRootLogger

[Neenu1995]

• Replaced String.format with StringBuilder and Guava's Strings for faster padding.
• Eliminated Thread.getAllStackTraces() (which triggered heavy JVM safepoints) by using Thread.currentThread() and Thread.enumerate() fallbacks.
• Cached PROCESS_ID instead of parsing it dynamically on every log line.
• Swapped in ThreadLocal<SimpleDateFormat> to safely reuse expensive date formatters.

[gemini-code-assist]

Code Review

This pull request refactors the BigQueryJdbcRootLogger to improve logging performance by replacing String.format with StringBuilder and introducing a ThreadLocal for SimpleDateFormat. It also changes how thread names and the process ID are retrieved. The review feedback identifies a potential NumberFormatException when parsing the process ID on certain JVMs, a regression in thread name lookup that fails to account for multiple thread groups, and suggests making the ThreadLocal date formatter static to optimize resource usage.

[logachev]

Maybe worth adding local cache for id->name?

[Neenu1995]

This fallback is an expensive, low-frequency edge case. The memory overhead of maintaining the cache exceeds the performance gains from faster lookups.