[NSE-856] Optimize of string/binary split

[FelixYBW]

What changes were proposed in this pull request?

manually split the string/binary column.
Previously we use builder to construct the string buffer. The PR create the destination buffers and read the src buffers, then split the column manually

The PR also removed some unused code

[github-actions]

#856

[FelixYBW]

split 5 columns of lineitem table, 3456 reducers on CLK machine

	1 thread	24 threads
base	6.98	17.4
c776c48	2.21	3.82

[FelixYBW]

Optimizations:

3456 partition	1 thread	24 thread
default	6.98	17.4
opt1	2.21	3.82
prefetch	1.94	3.71
add capacity check	2.12	3.9
avx cpy	1.92	4
16k batch	2.44	5.66
value_offset cached	2.18	3.33
remove partition_binary_buffer_idx_offset_	2.16	3.26
large page, 2M	2.08	2.92
allocate value buffer from pool	1.92	2.39
9 + prefetch dst next pid	1.85	2.43