Simplify br_if by removing its value operand.

[sunfishcode]

This proposes removing br_if's value operand (effectively by fixing the arity to 0) to defer the question of how its return value should be handled. We'd have the flexibility to reintroduce this operand in the future.

This new PR is in place of reopening #681 which github won't let me re-open.

[kripken]

I think this is a good option since

1. If we don't do this then we'd need to spend more time investigating (assuming the current data isn't convincing enough, which is fair).
2. Functionally, we can still do a conditional br with a value, using a br inside an if.

[ghost]

It's still not a great outcome as the 'arity' immediate is still there consuming space, just because a decision can not be made. Why not just propose to remove the 'arity' entirely, and if added in future then a separate opcode can just be added and we still have the more compact opcode for this high frequency case. Also the operator table has not been proposed yet, and there seems reasons to expect some objections particularly if it is a burden for a literal wasm interpreter.

[dschuff]

Still catching up after vacation, but I think this is a good idea too.

[sunfishcode]

br_if is being simplified via WebAssembly/spec#330 instead.