join_return_with_assignment tests should include assignment patterns

[srawlins]

Child issue of #58837

I think there should be test cases like:

var (a, b) = record; // variable declaration
return a;

// also
(a, b) = record; // just an assignment
return a;

in which the lint rule would not report that the return should be joined with the assignment.

(There are also no test cases with multiple assignment expressions in one statement, like

• a = b = 7; return b; or
• var a = 8, b = 9; return b;

)

[scheglov]

We don't report this lint even for a single declared local variable.

int f(int a) {
  var a = 0;
  return a;
}

We could, I guess, although this is not an assignment.

[scheglov]

As for multiple assignments, e.g.

int f1(int a, int b) {
  a = b = 0;
  return a;
}

int f2(int a, int b) {
  a = b = 0;
  return b;
}

Semantically it is not wrong to inline as return a = b = 0; in both cases.
But it definitely feels weird.
And we report the lint only for one of them, f1.

[srawlins]

Ah good analysis; we don't need to rock the boat for patterns; if declarations aren't covered, then they aren't covered. No biggie. But the assignment pattern (identifier pattern?) I think should still have a test.

[scheglov]

Tests were added.