refactor deferred balance to use memkv

[udpatil]

Describe your changes and provide context

This PR does the following:

• refactors the deferred logic to use a memKV store instead of an in memory map (allows us to avoid committing the changes when we discard other changes upon antehandler failure)
• refactors the feecollectors / deferred balance usage to key on module-txIndex-denom -> sdk.Coin
  • when writing the deferred balances, it first adds up all of the tx specific balances, then writes them in sorted order of the module addresses
• refactors the anteDependencyGenerator to take in a txIndex to allow for generating the dependencies appropriately for feecollector resources
• adds in a txIndex to sdk.Context to contain the currently executing TX index throughout delivertx context in order to allow for writing to the partitioned KV store to avoid resource overlap
• Adds a new instantiation function for BankKeeper to instantiate with the deferredCache initialized (deferred usages panic if not set up with this new function) (have to keep old one because wasmd breaks otherwise)
• Introduces the resource types specific to the deferredCache Store Key (because this is a separate KV store from bank store it has its own resource type and isn't under the Bank resource type

Still in progress:

• removing the context cache usages + deferred bank ops in memory structure
• adding unit tests for the memKV deferredCache that is part of bank keeper

Testing performed to validate your change

local sei testing + adding more unit tests

[codecov]

Codecov Report

Merging #287 (e5067f1) into main (db47ec3) will decrease coverage by 0.25%.
The diff coverage is 67.60%.

❗ Current head e5067f1 differs from pull request most recent head 21c819d. Consider uploading reports for the commit 21c819d to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #287      +/-   ##
==========================================
- Coverage   55.17%   54.93%   -0.25%     
==========================================
  Files         619      618       -1     
  Lines       51569    51531      -38     
==========================================
- Hits        28455    28310     -145     
- Misses      21063    21169     +106     
- Partials     2051     2052       +1     

Impacted Files	Coverage Δ
baseapp/baseapp.go	72.32% <0.00%> (-1.43%)	⬇️
types/accesscontrol/comparator.go	75.67% <ø> (ø)
types/accesscontrol/resource.go	0.00% <ø> (ø)
x/accesscontrol/types/graph.go	77.82% <ø> (ø)
x/auth/ante/basic.go	69.91% <0.00%> (ø)
x/auth/ante/sigverify.go	63.35% <0.00%> (ø)
x/bank/types/key.go	35.13% <0.00%> (-23.96%)	⬇️
types/handler.go	58.92% <8.33%> (ø)
types/context.go	73.88% <52.94%> (-0.15%)	⬇️
x/bank/keeper/deferred_cache.go	77.33% <77.33%> (ø)
... and 4 more

... and 4 files with indirect coverage changes

[udpatil]

cc @BrandonWeng is there a specific reason you have previously called the dependency generator here? I assume it was because you may not have been aware that the -1 key was reserved for the access ops generated for the ante handler ahead of time? but just want to make sure there wasnt some other reason

[BrandonWeng]

Teah that was most likely the case

[alexanderbez]

nit: What is prior to messages in the transaction?

[udpatil]

the ante handler, since that also needs to declare resource usages. When constructing the DAG, we generate the dependencies on a per message basis, so for simplicity we treat the antehandler as the -1st message in the list of messages

[alexanderbez]

Sorry, I meant the godoc should be updated to reflect that :p

[udpatil]

ah yeah, will do

[BrandonWeng]

not a blocker for this PR but could we add some unit tests for the error cases? the happy paths are tested quite well already

[BrandonWeng]

nit but could we add unit tests for failure scenarios? The happy paths seem to be covered quite well already. Not a blocker for this PR