Refactor block type restriction underscore output to nil

[straight-shoota]

This patch replaces underscore output types in block restrictions with nil (empty) for methods where the result of the yielded block is irrelevant and unused. Underscore should be used where the result type is not specified but the value is actually used.

See #10931

This had previously been a part of #10929

[beta-ziliani]

I wonder if it would break things to explicit the Nil return type when that's the intention (e.g., each)

[straight-shoota]

Nil and empty should be semantically equivalent here. We could use Nil, it's just more characters.

[beta-ziliani]

right, but given that the semantics are somewhat not entirely there, maybe it makes sense to be explicit for once

[straight-shoota]

Let's discuss the intended semantics in #10931

I just opened this PR to not loose the these parts of #10929. It should have been a draft from the beginning, to wait for a definitive decision in #10931.