[improvement] Consider remove common/allocators

[BohuTANG]

Summary

Memory allocator biggest improvement for Datafuse is to make block allocation.
For example, it will have a big performance gain when using snmalloc, since Datafuse has many pieces allocation, such as Vec<T> and many points not optimized yet.

But in the latest version, Datafuse has many improvements(#1444 #1334 #1239 #1214) about the memory and the algorithm, less and less reliance on the memory allocator.
From my test:

1. With snmalloc

Hardware: AMD Ryzen 7 PRO 4750U, 8 CPU Cores, 16 Threads
2021-08-15T05:30:42Z INFO  datafuse_query] DatafuseQuery v-0.1.0-f3780e1-simd(1.56.0-nightly-2021-08-15T04:19:45.246023742+00:00), Allocator: snmalloc

clickhouse-client --time --multiquery --format Null <<<"
    SELECT avg(number) FROM numbers_mt(100000000000);
    SELECT sum(number) FROM numbers_mt(100000000000);
    SELECT min(number) FROM numbers_mt(100000000000);
    SELECT max(number) FROM numbers_mt(100000000000);
    SELECT count(number) FROM numbers_mt(100000000000);
    SELECT sum(number+number+number) FROM numbers_mt(100000000000);
    SELECT sum(number) / count(number) FROM numbers_mt(100000000000);
    SELECT sum(number) / count(number), max(number), min(number) FROM numbers_mt(100000000000);
    SELECT number FROM numbers_mt(10000000000) ORDER BY number DESC LIMIT 10;
    SELECT max(number), sum(number) FROM numbers_mt(1000000000) GROUP BY number % 3, number % 4, number % 5 LIMIT 10;"
5.763
6.038
6.236
6.285
5.686
16.614
6.171
8.970
2.612
1.836

2. Rust default allocator

Hardware: AMD Ryzen 7 PRO 4750U, 8 CPU Cores, 16 Threads
[2021-08-15T05:32:28Z INFO  datafuse_query] DatafuseQuery v-0.1.0-f3780e1-simd(1.56.0-nightly-2021-08-14T15:03:17.462456200+00:00), Allocator: default

clickhouse-client --time --multiquery --format Null <<<"
    SELECT avg(number) FROM numbers_mt(100000000000);
    SELECT sum(number) FROM numbers_mt(100000000000);
    SELECT min(number) FROM numbers_mt(100000000000);
    SELECT max(number) FROM numbers_mt(100000000000);
    SELECT count(number) FROM numbers_mt(100000000000);
    SELECT sum(number+number+number) FROM numbers_mt(100000000000);
    SELECT sum(number) / count(number) FROM numbers_mt(100000000000);
    SELECT sum(number) / count(number), max(number), min(number) FROM numbers_mt(100000000000);
    SELECT number FROM numbers_mt(10000000000) ORDER BY number DESC LIMIT 10;
    SELECT max(number), sum(number) FROM numbers_mt(1000000000) GROUP BY number % 3, number % 4, number % 5 LIMIT 10;"
3.301
3.177
4.301
4.302
2.203
16.466
3.680
8.047
3.140
1.764

It is only advantageous in the case of ORDER BY , so there should still be some fragmented memory allocation in ORDER BY , this can be optimized in future.
Rust default memory alloctor is good for Datafuse now.

[BohuTANG]

cc: @sundy-li @zhang2014

[zhang2014]

Our biggest motivation for expanding the memory allocator may be to track the allocation of memory #1148

[BohuTANG]

Our biggest motivation for expanding the memory allocator may be to track the allocation of memory #1148

Yes, it will be another issue