Binary Reading: Don't pick internal names for things that overlap with the name section

[kripken]

This can only really happen in the fuzzer, but it is annoying there:
we can have fimport$17 in the names section, and before this
fix, we could pick that as the implicit name for something without
a name section entry. The fix is just to note names from the
names section in the list of used names.

[tlively]

Is there anything stopping us from generating a name below and later running into it in the names section?

[kripken]

I believe this code aims to avoid that, be pre-populating the names:

binaryen/src/wasm/wasm-binary.cpp

Lines 3010 to 3020 in efcf2dc

	void WasmBinaryReader::readFunctionSignatures() {
	size_t num = getU32LEB();
	auto numImports = wasm.functions.size();
	std::unordered_set<Name> usedNames;
	for (auto& [index, name] : functionNames) {
	if (index >= num + numImports) {
	std::cerr << "warning: function index out of bounds in name section: "
	<< name << " at index " << index << '\n';
	}
	usedNames.insert(name);
	}

But that raises the question of why that didn't fix this issue. Unfortunately, I can only reproduce this with an experimental pass not yet in a PR, so I don't have a good testcase. Anyhow, let's hold off on this until that PR is ready.