Turn shl+shr of the same amount into a mask

[kripken]

E.g.

(i32.shl
 (i32.shr_u
  (local.get $0)
  (i32.const 4)
 )
 (i32.const 4)
)
 =->
(i32.and
 (local.get $0)
 (i32.const -16) ;; 0xfffffff0
)

We optimize the reverse, x << 16 >> 16 becomes a sign-extend, but not this.

Found by the superoptimizer #4994 (for comparison to other findings: rule #28, benefit 11044).

[MaxGraey]

Already made PR for this: #4996

[kripken]

Thanks @MaxGraey ! Reading now.