What problem does this PR solve?

Issue Number: ref #9032

Problem Summary:

What is changed and how it works?

Depends on #9678, #9690

1. support filter in deserializeBinaryBulk so that we can only deserialize necessary rows
2. refactor DMFileReader to pre-compute read block information so that we can know the block size before reading.
3. refactor vector search path to use readWithFilter rather than read when there are vector indexes.

This PR can improve 8.5% QPS of vector search in scenarios with updates.

before

"metrics": {
    "qps": 114.8094,
    "serial_latency_p99": 0.0876,
    "recall": 0.9324
}

after

"metrics": {
    "qps": 124.6441,
    "serial_latency_p99": 0.0815,
    "recall": 0.9315
}

Why it has no benefit in scenarios with static data?

1. When data-sharing is enabled, always deserialize all rows.
2. There is a PK column cache in DMFile, there is no need to read the PK column in DMFile.

Improve 8.5% the performance of vector search in scenarios with updates.

Check List

Tests

• Unit test
• Integration test
• Manual test (add detailed scripts or steps below)
• No code

Side effects

• Performance regression: Consumes more CPU
• Performance regression: Consumes more Memory
• Breaking backward compatibility

Documentation

• Affects user behaviors
• Contains syntax changes
• Contains variable changes
• Contains experimental features
• Changes MySQL compatibility

Release note

None