Go: Fix data flow through variable defined in type switch guard

[owen-mc]

Reported here.

Consider a typical type switch statement.

	switch inputWithType := input.(type) {
	case Type1:
		output1 = f(inputWithType)
	case Type2:
		output1 = g(inputWithType)
	}

It looks like a new variable inputWithType is defined in the guard, and is used in the case clauses. But the go compiler can't represent it like that internally because inputWithType has a different type in each case clause. So it has to make a variable for each case clause. The CodeQL Go library is not currently dealing with this. We end up with an object for inputWithType which doesn't have a declaration, which means that there is no SsaDefinition relating to it. This causes data flow to not work.

I tried fixing this in the extractor, but it was a bit messy, and I realised it was better to accept that these variables don't have an explicit declaration and add an IR::Instruction for their implicit declaration. This causes SSA to work properly again. I have added tests for control flow and data flow.

[smowton]

Liking that we apparently don't even need to touch the SSA layer -- this becomes just another instruction that defines a fresh variable 👍

[owen-mc]

I found a way to identify the implicitly declared objects without a new table. This simplifies the PR a bit.

[smowton]

QL changes look good; trivial suggestion re: change note. Haven't examined the test expectations in detail.

[owen-mc]

QA results all look good.