[SPARK-45625][BUILD] Upgrade log4j to 2.21.0

[LuciferYang]

What changes were proposed in this pull request?

This pr aims upgrade log4j from 2.20.0 to 2.21.0.

Why are the changes needed?

Support for the zstd compression algorithm has been added in the new version: apache/logging-log4j2#1508 | apache/logging-log4j2#1514
Meanwhile, the new version starts to use Java 11 for building, and the runtime version is still compatible with Java 8: apache/logging-log4j2#1369
The new version also brings some bug fixes, such as:

• Fixed logging of java.sql.Date objects by appending it before Log4J tries to call java.util.Date.toInstant() on it: Error logging java.sql.Date logging-log4j2#1366
• Fixed concurrent date-time formatting issue in PatternLayout: Fix concurrent date-time formatting issue in PatternLayout logging-log4j2#1485
• Fixed buffer size in Log4jFixedFormatter date time formatter: ArrayOutOfBoundsException in Log4jFixedFormatter logging-log4j2#1418
• Fixed the propagation of synchronous action failures in RollingFileManager and FileRenameAction: Exceptions are ignored even with ignoreExceptions="false" when booting logging-log4j2#1445 | Issue 1445 - Fix RollingFileManager and FileRenameAction logging-log4j2#1549
• Fixed RollingFileManager to propagate failed synchronous actions correctly: Exceptions are ignored even with ignoreExceptions="false" when booting logging-log4j2#1445
  and more.

The complete release note is as follows:

• https://github.com/apache/logging-log4j2/releases/tag/rel%2F2.21.0

Does this PR introduce any user-facing change?

No

How was this patch tested?

Pass GitHub Actions

Was this patch authored or co-authored using generative AI tooling?

No

[LuciferYang]

Merged into master for Spark 4.0. Thanks @HyukjinKwon