Add feature for tracking dynamic reserved account set

[jstarry]

Problem

There's no way to add new builtin programs and sysvars to the list of read-only reserved accounts.

Summary of Changes

Bank now tracks a dynamic set of reserved account keys which gets updated on epoch boundaries with feature activations. The set of reserved account keys needs to get passed to all the transaction and message constructors and then we take advantage of the is_writable cache to avoid passing the reserved account set to is_writable everywhere. It's inevitable that a change like this will result in breaking changes unfortunately. There is an option of moving towards using more internal sdk crates (like solana-runtime-transactions) for core types rather than thrashing the external sdk whenever internal needs change, but that's a pretty big effort.

Detailed changes

1fa7b7d: refactors tests to minimize diff size when plumbing reserved accounts
c8cb15f: adds the zk token module to the sdk so that reserved accounts can reference the program id
9ea7043: removes the TryFrom<Message> impl for SanitizedMessage and replaces it with SanitizedMessage::try_from_legacy_message which will allow plumbing the reserved accounts through
edbdf6f: adds Message::is_maybe_writable for times when the reserved accounts set is not available but we want a rough guess still (cli output, durable nonce checking in rpc, etc)
8fb6616: is the meat of the change, it introduces the reserved accounts module, adds it to bank, and plumbs the reserved accounts set everywhere
701119b: deprecates uses of static lists of sysvars and builtins. Unfortunately the lists themselves can't be deprecated due to use of lazy_static

• Deprecated solana_program::message::is_builtin_key_or_sysvar function
• Deprecated solana_program::sysvar::is_sysvar_id function
• Bigtable uploader used to skip indexing sysvars, now it skips all reserved keys
• Replaced some usages of is_sysvar_id which a check for owner equal to sysvar
• Snapshot minimizer logic now adds all active and inactive reserved account keys to the minimized snapshot

Breaking changes:

• Renamed solana_rpc_client::rpc_client::SerializableTransaction::uses_durable_nonce to maybe_uses_durable_nonce to prevent misuse
• Renamed solana_sdk::transaction::uses_durable_nonce to maybe_uses_durable_nonce to prevent misuse
• Renamed solana_sdk::transaction::VersionedTransaction::uses_durable_nonce to maybe_uses_durable_nonce to prevent misuse
• solana_sdk::transaction::SanitizedTransaction::try_from_legacy_transaction has a new reserved_account_keys argument
• solana_sdk::transaction::SanitizedTransaction::try_create has a new reserved_account_keys argument
• solana_sdk::transaction::SanitizedTransaction::try_new has a new reserved_account_keys argument
• solana_program::message::Message::is_writable has a new reserved_account_keys argument
• solana_program::message::Message::get_account_keys_by_lock_type has a new reserved_account_keys argument
• solana_program::message::LegacyMessage::new has a new reserved_account_keys argument
• solana_program::message::SanitizedMessage::try_new has a new reserved_account_keys argument
• solana_program::message::LoadedMessage::new has a new reserved_account_keys argument
• solana_program::message::LoadedMessage::new_borrowed has a new reserved_account_keys argument
• solana_program::message::SanitizedMessage no longer implements TryFrom<Message> (I added a try_from_legacy_message method instead in order to pass the reserved_account_keys argument)

Feature Gate Issue: #34899

[t-nelson]

oof. the number of touch points here begs a refactor... i'm going to have to look at it more closely 🤔

[jstarry]

I don't disagree that a refactor of transaction / message types should be done at some point but I would like to point out that the plumbing for reserved_account_keys is pretty similar to how address lookup tables are loaded since both are bank specific. I think the current approach in this PR already minimizes touch points as much as is possible.

[codecov]

Codecov Report

Attention: Patch coverage is 91.09415% with 35 lines in your changes are missing coverage. Please review.

Project coverage is 81.7%. Comparing base (bf2e8ee) to head (ea9e3ca).
Report is 6 commits behind head on master.

❗ Current head ea9e3ca differs from pull request most recent head 9f6432b. Consider uploading reports for the commit 9f6432b to get more accurate results

Additional details and impacted files

@@            Coverage Diff            @@
##           master   #34901     +/-   ##
=========================================
- Coverage    81.7%    81.7%   -0.1%     
=========================================
  Files         834      830      -4     
  Lines      224232   224570    +338     
=========================================
+ Hits       183382   183476     +94     
- Misses      40850    41094    +244     

[CriesofCarrots]

Is it possible to break this up into logical commits? Right now, review is horrible.

[jstarry]

Is it possible to break this up into logical commits? Right now, review is horrible.

I split it up.. apologies for not doing that earlier! The PR description includes a detailed breakdown of each commit as well.

[jstarry]

Would you prefer I open a pr for each individual commit? Also, considering that 2.0 is coming soon, breaking changes should be an issue?

[jstarry]

@CriesofCarrots @t-nelson ping on this!

[CriesofCarrots]

Sorry for the delay. This is on my list for tomorrow. Is it worth waiting on a rebase at this point?

[jstarry]

@CriesofCarrots I just rebased!

[CriesofCarrots]

Phew, okay, I think I got through everything. (Commit 5 was a bear; splitting out test fixes from other updates might have been nice to draw attention to the critical code paths.)
Generally, looks good to me; just one api suggestion. I didn't catch any issues with the feature/list handling.
The banking_stage changes look straightforward to me, but it might be worth having an module expert sanity check.

[CriesofCarrots]

lgtm, but recommend at least one more set of eyes on everything

[willhickey]

This repository is no longer in use. Please re-open this pull request in the agave repo: https://github.com/anza-xyz/agave