lang: adjust realloc implementation to safeguard max increase and idempotency

[callensm]

• includes the original implementation of the realloc constraint group from lang: add realloc constraint group #1943
• adds a check on the __delta_bytes to ensure it is <= MAX_PERMITTED_DATA_INCREASE
• add test to ensure duplicate accounts does not imply duplicate reallocation
• adds new AccountReallocExceedsLimit error code to rust and typescript for the data increase limit check

[callensm]

added reallocs: &mut std::collections::BTreeSet<String> as a new argument to Accounts::try_accounts in order to track public keys that are assigned to reallocation constraint groups in order to throw a duplicate account realloc error if the public key of an account is associated with more than one reallocation per instruction call.

[ethanwu10]

It seems that coverage of #[state] is lacking for zerocopy. Also, the second typo in the ix.has_receiver case for state interface impls is dead (and thus also untested), since #[interface] does not allow receivers in any of the trait's methods.

Also, I didn't see anything in rustdoc indicating that #[state] is deprecated...

[callensm]

tests/swap is passing when i run it locally with the debug build of the cli, not sure why its failing in the pipeline.

[ethanwu10]

Since AFAICT nothing in docs signals that the Accounts trait is internal, this is a breaking change (with no real way around it if we want to implement the duplicate checking). It'd probably be a better idea to pass a new &mut TryAccountsContext struct or similar so that we can add more bits of state or data without needing to further modify this signature. Consider:

#[derive(Default)]
pub struct TryAccountsContext {
    // private field to prevent the struct from containing only public fields; prevents users from
    // constructing using a literal or pattern-matching/destructuring against this type
    _marker: (),
    pub bumps: BTreeMap<String, u8>,
    pub reallocs: BTreeSet<String>,
}

[vercel]

@armaniferrante is attempting to deploy a commit to the 200ms Team on Vercel.

A member of the Team first needs to authorize it.

[edmbn]

added reallocs: &mut std::collections::BTreeSet<String> as a new argument to Accounts::try_accounts in order to track public keys that are assigned to reallocation constraint groups in order to throw a duplicate account realloc error if the public key of an account is associated with more than one reallocation per instruction call.

Will this mean we won't be able to reallocate an account 2 times in the same instruction? We might want to reallocate an account with 0 space and then with try_accounts reallocate to original space to have the account "reinitialized".