[wip] bloatnet: reduce word-level compress sampling

[AskAlexSharov]

to speedup merge: increase sampling of DomainCompressCfg and HistoryCompressCfg
@sudeepdino008 do you thing we must go this way? Need cherry-pick some bug-fix?

[sudeepdino008]

tldr:

• for bloatnet, this PR is okay - it'll speedup merge (-30%) and the file is slightly smaller (db/state: raise domain compression SamplingFactor to 4 #21639 (comment)); there is a bug but it doesn't matter for correctness and even if it is fixed, the merge will still be faster than today and filesize similar. So this PR is good for bloatnet
• mainnet: scenario is different -- probably because of higher entropy data -- I saw 43% increase in size (merge was 5x faster though);
  • after the fix merge is only 2x faster, but 13% bigger files.

---

• So this PR "works for bloatnet, but not for mainnet"; -- I'd suggest to go ahead with it on bloatnet, maybe put it in "bloatnet specific args".
• the algorithm change - [wip] seg: 2-3x faster domain merge compression via Aho-Corasick matcher and cover-DP optimizations #21625 -- saw 2x speedup on bloatnet -- it worked in my testing (byte identical etc.) -- but I don't want to push this version in main atleast; in performance -- maybe we can do it (upto you)
• acceptable for mainnet too but only with the bug fix.
• I don't have stats for history merge...but definitely will be faster.

[AskAlexSharov]

@sudeepdino008 seems you saw on mainnet +13% size on SF=4. Right? (not +43%)

[sudeepdino008]

@sudeepdino008 seems you saw on mainnet +13% size on SF=4. Right? (not +43%)

with the #21628 fix, it was 13%, without the fix 43%.
I'll add the fix into performance