Fix schema evolution gaps for FG reader

We tested schema evolution support from FG reader through three tests:

CASE1: Two base files have different schemas.

CASE2: One base files with multiple log files, and log files have different schemas.

CASE3: One base file and one log file have different schemas.

Using the above tests, we found that there are some failures for:
# Adding nested column for Hive: CASE1, CASE2 
# Adding nested column for Java: CASE2
# Type upgrade from int to long Hive: CASE2
# All test cases failed for Flink: probably due to some bugs in the test.

The PR for running tests: [https://github.com/apache/hudi/pull/13443]

## JIRA info

- Link: https://issues.apache.org/jira/browse/HUDI-9561
- Type: Improvement
- Fix version(s):
- 1.1.0