Fix - bpf_account_data_direct_mapping

[Lichtso]

Problem

Summary of Changes

This PR has a carefully curated commit history and the commit messages explain each refactoring step. Please review commit by commit in order.

Open TODOs for the new account_data_write_access_handler:

• It is incorrect because in a rare reallocation pattern it can result in a dangling pointer which causes the test EXTEND_AND_WRITE_U64 of test_program_sbf_realloc to be flaky.
• It should be optimized further to only allocate the realloc region if necessary, not if only the account data is touched.
• It should be optimized even further by extending the account region into the virtual address space occupied by the realloc region so that each account only has one region carrying its payload and the realloc region is always uninitialized or non existent.
• For all of the above it would need access to multiple memory regions at once and be region independent by getting a virtual address as parameter.

Open TODOs for the logic which decides when and how much Vec capacity to zero:

• We might want to stop abusing the Vec::capacity and instead track an explicit additional field for each account.
• The buggy way the realloc detection works needs to be completely rewritten.

There is a new behavioral change in this patch, which I don't expect programs to depend on, but it needs to be tested / measured on MNB. See:

• SIMD-0219
• Adjustment of TEST_CPI_ACCOUNT_UPDATE_CALLER_GROWS_CALLEE_SHRINKS.

Feature Gate Issue: https://github.com/anza-xyz/feature-gate-tracker/issues/16

[mergify]

The Firedancer team maintains a line-for-line reimplementation of the
native programs, and until native programs are moved to BPF, those
implementations must exactly match their Agave counterparts.
If this PR represents a change to a native program implementation (not
tests), please include a reviewer from the Firedancer team. And please
keep refactors to a minimum.