Make sure HoodieTableMetadata API allows to read both in-memory and on-cluster

Currently, most of the HoodieTableMetadata APIs (getColumnStats, getBloomFilters, etc) provide only for loading the data from MT in-memory which shifts the burden on the caller to break their requests in chunks to make sure these fit in memory: for ex, trying to read Bloom Filters from MT, Bloom Index (caller) have to make sure we're not reading more than 256 filters at a time to limit its memory footprint.

Instead HoodieTableMetadata API has to be
# Rebased to rely on HoodieData
# Provide levers to the caller whether MT should be read in-memory or on-cluster
#  

## JIRA info

- Link: https://issues.apache.org/jira/browse/HUDI-5556
- Type: Improvement
- Epic: https://issues.apache.org/jira/browse/HUDI-1292
- Fix version(s):
- 1.1.0