Add function alignment for better performance

[danlark1]

I hereby agree to the terms of the CLA available at: https://yandex.ru/legal/cla/?lang=en

Changelog category (leave one):

• Performance Improvement

Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):
Add function alignment for possibly better performance.

Detailed description / Documentation draft:
Let's try to add compression of debug symbols and reduce the binary size by 1.2GB. Also, adding the function alignment for more reproducible results in the performance tests. It is a known technique to reduce such randomness and should potentially help. Either way, let's try and see the build and performance metrics. falign-functions add 0.1% to the binary size which is nothing compared to other clickhouse

[azat]

Note that this will break path/line numbers in stacktrace (as pointed by @alexey-milovidov in #18952 (comment))

[alexey-milovidov]

@azat Need to check it - maybe it started to work.

[alexey-milovidov]

For function alignment - the changes in performance are unclear, maybe it is faster for 0.3% in average. Good to merge.
For changes in stability of perf tests, we should compare two new versions with this change (after merge).

[alexey-milovidov]

Integration tests (thread) — fail: 1, passed: 1040, error: 24

Broken in master.

[alexey-milovidov]

ClickHouse special build check — 5/6 builds are OK

Need to disable the flag for Darwin build.

[azat]

@azat Need to check it - maybe it started to work.

@alexey-milovidov yep, I've re-checked before writing, it still does not work.

[alexey-milovidov]

Ok, I removed compression as we cannot use it for now.
Let's validate the changes in performance.

Previous run showed -0.3% in average, it can be a noise.

[alexey-milovidov]

There is zero difference in performance. Also looks like no difference in test stability.
There are no queries with statistically significant speedup.

[danlark1]

Then I am ok not to merge it for now

[alexey-milovidov]

Note that all performance tests are performed on Xeon Gold CPUs. Maybe there is some difference on other CPUs but we don't know.

@danlark1 do you think we should merge it?

[alexey-milovidov]

Ok.

[alexey-milovidov]

@danlark1 said that it just should be better.