Change fuzzbench_qemu fuzzer

[tokatoka]

• don't test libpng
• instead run the same harness as fuzzbench to see if it find objectives or not
• added these too to self-hosted ci

but it doesn't work as I said in discord. this is probably due to signal handling is broken in Qemu?

also fuzzbench_fork_qemu doesn't seem to detect any new edges. can you check that too when you are back? @rmalmain

[tokatoka]

I forgot to say this. but fuzzbench_fork_qemm doesn't seem to be correctly instrumented @rmalmain

[rmalmain]

yes i realized with the ci, i am still debugging this one. fuzzbench_qemu test works correctly locally at least now.

[rmalmain]

i fixed fork executor and reduced some trait bounds as well. should be good to go, works locally

[tokatoka]

ok ok Rust toolchanins are broken as always!! 👎 😠

[tokatoka]

qemu fuzzer still reporting errors

[rmalmain]

yes it's normal, i also fixed the fork executor harness' signature

[tokatoka]

depends on #2496

[rmalmain]

few things:

• renaming of EDGES_MAP_SIZE_IN_USE into EDGES_MAP_DEFAULT_SIZE and EDGES_MAP_SIZE_MAX into EDGES_MAP_ALLOCATED_SIZE.
• EdgeCoverageModule stuff can only be built with a reference to observer through the builder. This is a first step to make the module work with different kinds of MapObserver
• integration of VariableLengthMapObserver and ConstantLengthMapObserver traits to be able to differentiate both map styles at compile time. for now i only implemented VariableLengthMapObserver for testing.

[rmalmain]

what you think of the changes @tokatoka ? does smth look bad?

[tokatoka]

it's good

[tokatoka]

@rmalmain there're python errors to fix

[tokatoka]

also can you fix the fuzzer examples?

[rmalmain]

TODO: Implement EdgeModule with ConstantLengthMapObserver variant