fix: Detect ABI change of fuzzing harnesses

[Rumata888]

Description

Problem*

Changing a fuzzing harness with an existing corpus could lead to panics, if former elements couldn't be cast to appropriate types.

Summary*

Before fuzzing with the starting corpus we now try to decode all testcases with the current ABI. If a testcase fails, then we collect parse it for dictionary elements.
For example, if you first fuzz

#[fuzz(should_fail)]
fn fuzz_main(x: Field, y: u64, z: u64) {
    
     let prod=y*z;
     assert(y<0x100000000);
     assert(z<0x100000000);
     assert(prod-0xdeadbeefdeadbeef<0xdeadbeef);
     assert(x==0xdeadbeefdeadbeefdeadbeefdeadbeef);
 }

and find a failure (nargo fuzz), after changing to

#[fuzz(should_fail)]
fn fuzz_main(x: u64, y: u64, z: u64) {
    
    let prod=y*z;
    assert(x<0x100000000);
    assert(y<0x100000000);
    assert(z<0x100000000);
    assert(prod-0xdeadbeefdeadbeef<0xdeadbeef);
}

The fuzzer will inform you that the ABI has changed and will almost immediately find a failing testcase, because it already has useful values that it needs to put into y and z

Additional Context

Documentation*

Check one:

• No documentation needed.
• Documentation included in this PR.
• [For Experimental Features] Documentation to be submitted in a separate PR.

PR Checklist*

• I have tested the changes locally.
• I have formatted the changes with Prettier and/or cargo fmt on default settings.

[asterite]

Looks great!