fix: collect push bytes from test setup in fuzzer

[onbjerg]

Motivation

There's an old report that immutables were not collected in the fuzzer. Immutables are sections in the bytecode of contracts that are replaced by constant values, so we should be collecting them (since we collect push bytes), but it turns out that we were not collecting push bytes in the initial fuzzer DB (i.e. from the state right after setUp).

See #1168

Solution

Also collect push bytes after setUp. Closes #1168

Note: The state dict after setUp for the test in #1168 contains about ~120 values. It seems that either one of the tests fail every time (which is what we want), but not super reliably. Increasing the fuzzer runs to 512 makes them both fail reliably, though.

[mattsse]

smol nit

[mattsse]

unsure if this is related but this now results in

Failing tests:
Encountered 2 failing tests in src/test/StdAssertions.t.sol:StdAssertionsTest
[FAIL. Reason: Too many global rejects] testAssertEq_BytesErr_Pass(bytes,bytes) (runs: 31, μ: 20023, ~: 21439)
[FAIL. Reason: Too many global rejects] testAssertEq_Bytes_Pass(bytes,bytes) (runs: 24, μ: 19965, ~: 20697)

when running forge test on forge-std, I could reproduce this locally

[joshieDo]

We had this discussion here: #2724 (comment)

We decided to revert the push collection on build_initial_state. But I'm all for keeping this and change the forge-std test instead.

[onbjerg]

The Forge Std assume call is just way too narrow so I'd expect it to reach the limit, and the fact that the reject limit hasn't been reached before is not ideal. Either they can increase the reject limit in Forge Std or just fix the test, obviously if the assume call makes it so a is always b then they should just take a single parameter.

[mds1]

Just merged foundry-rs/forge-std#160 which should prevent foundry CI from failing on those now

[gakonst]

Rebased, merging per the forge-std fixes. Flagging for us to keep this PR in mind if people come to us with complaints about the fuzzer error'ing too much.