remove Dockerfile, use only pure go dabase backends

[faddat]

This PR:

• updates golang to v1.22.* and pebble v1.1.0
• allows cometbft downstream to reduce its usage of Docker
• reduces CI time on this repository by over 95%
• keeps goleveldb for backwards compatiblity, and badger because it's pure go
• eliminates unmaintained database backends with no future:
  • bolt
• eliminates database backends that complicate both cometbft-db and cometbft by forcing the use of cgo and docker:
  • rocksdb / grocksdb
  • cleveldb
• eliminates the need for build tags

[codecov]

Codecov Report

Attention: Patch coverage is 50.00000% with 2 lines in your changes are missing coverage. Please review.

Project coverage is 77.26%. Comparing base (fb4f703) to head (e3118b3).
Report is 11 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #135      +/-   ##
==========================================
+ Coverage   76.80%   77.26%   +0.45%     
==========================================
  Files          23       14       -9     
  Lines        2048     1060     -988     
==========================================
- Hits         1573      819     -754     
+ Misses        403      193     -210     
+ Partials       72       48      -24     

Files	Coverage Δ
pebble.go	72.64% <100.00%> (-0.92%)	⬇️
badger_db.go	89.03% <33.33%> (+1.25%)	⬆️

... and 11 files with indirect coverage changes

Files	Coverage Δ
pebble.go	72.64% <100.00%> (-0.92%)	⬇️
badger_db.go	89.03% <33.33%> (+1.25%)	⬆️

... and 11 files with indirect coverage changes

[faddat]

@melekes -- this is my new proposed PR. It is simpler, easier to maintain, and faster all around.

It will also reduce dramatically the amount of work needed to maintain both this repository and cometbft.

The correct answer to your question about the Dockerfile, is that there's been no reason to have a Dockerfile in this repository for years.

[faddat]

When this PR started, tests took more than ten minutes. They now take less than one minute.

[faddat]

I think I just fixed that annoying codeql thing.

[melekes]

eliminates unmaintained database backends with no future:
bolt
badger

badger is very much maintained https://github.com/dgraph-io/badger

[faddat]

@melekes it is pure go, too, so would you like to keep it?

[faddat]

I think this is ready to rock now, but there is a valid question of weather or not to include badger.

Badger v4 worked out of the box. No valid reason not to include it.

[faddat]

If we are going to merge this, we should. Blocks upgrading comet to go 1.22

[adizere]

I would prefer to sunset database backends iteratively:

1. mark them as deprecated -- comet-db v0.12
   • then give it some time for users to acclimatize
2. remove them -- comet-db v0.13

I would also keep rocksdb around, since some users (at least Cronos, potentially others) clearly make use of it.

Therefore, we should separate this PR into multiple smaller PRs:

• golang to v1.22.* and pebble v1.1.0
• mark cleveldb and bold as deprecated
• Docker improvements

@faddat do you plan on helping with any of the above? if so, please open corresponding PRs. if not, we plan to work on it this week so we'll go ahead within a couple of days

[faddat]

@adizere

it is necessary to update the dockerfile in order to update grocksdb. So ci will take a very long time now.

As for phases, that's fine.

We will not be able to improve docker without removing rocksdb (and other databases that use cgo)