[core] DataEvolution table supports concurrent updates to different columns

[steFaiz]

Purpose

Currently, if we concurrently trigger columns update jobs for data evolution table, the later jobs will be failed:

This happens even we are updating different columns.
However, we could only forbid updating the same columns. This PR introduces a RowIdColumnConflictChecker and use binary search to accelerate check process.

BenchMark

I tested 100000 files with disjoint row range check against 100000 files with disjoint row range, each file contains 3 columns, the result is:

Iteration	Check Files	Expected Merged Ranges	Check Time
0	100,000	100,000	11 ms
1	100,000	100,000	8 ms
2	100,000	100,000	8 ms
Best	100,000	100,000	8 ms
Average	100,000	100,000	9 ms

with binary search, the check cost is negligible.

Tests

UT Cases for Spark and Unit Tests for core.