fix: bitvec simproc bug (#14370)

This PR fixes a bug in the `BitVec` simproc when `bitVecOfNat := false`.
This bug
affects `grind` since it uses `bitVecOfNat := false`. Here is an example
reported by Henrik
that exposed the issue.

```lean
example (x : BitVec 4) (_h1 : x = 0#2 ++ 1#2) (_h2 : x = 1#4) : True := by
  grind
```

It also removes a redundant `grind` theorem, and adds support for
normalizing terms such as `0#n` when `bitVecOfNat := false`.
4 files changed