The implementation of `BooleanMonomialVariableIterator` is a bit slow.
It is based in polybori's `PBMonomVarIter` which implements an iterator
over the variables of a boolean monomial which is slow.

However there is `PBMonomIter` which implements an iterator over the
indices of the variables which seems to be faster; the variables can
then be retrieved from the ring itself.

This is what we do in this commit. Performance is way better.

I don't know why this difference, maybe because the variables are strings
and this suffers from allocation, while indices are just 32 bit ints.

Before this commit:
```
sage: R = BooleanPolynomialRing(512)
sage: f = sum(R.gen(randrange(512)) for _ in range(20))
sage: timeit('vs = f.variables()')
125 loops, best of 3: 6.1 ms per loop
```

After this commit:
```
sage: R = BooleanPolynomialRing(512)
sage: f = sum(R.gen(randrange(512)) for _ in range(20))
sage: timeit('vs = f.variables()')
625 loops, best of 3: 13.7 μs per loop
```

Another example taken from a doctest for the method
`PolynomialSequence_generic.connected_components()` in
`src/sage/rings/polynomial/multi_polynomial_sequence.py`:

Before this commit:
```
sage: sr = mq.SR(2,4,4,8,gf2=True,polybori=True)
sage: F, s = sr.polynomial_system()
sage: Fz = Sequence(F.part(2))
sage: %time vs = Fz.variables()
CPU times: user 603 ms, sys: 2.42 s, total: 3.03 s
Wall time: 3.03 s
sage: %time len(Fz.connected_components())
CPU times: user 2.59 s, sys: 10.9 s, total: 13.5 s
Wall time: 13.5 s
4
```

After this commit:
```
sage: sr = mq.SR(2,4,4,8,gf2=True,polybori=True)
sage: F, s = sr.polynomial_system()
sage: Fz = Sequence(F.part(2))
sage: %time vs = Fz.variables()
CPU times: user 9.77 ms, sys: 994 µs, total: 10.8 ms
Wall time: 13.7 ms
sage: %time len(Fz.connected_components())
CPU times: user 70.8 ms, sys: 8.27 ms, total: 79.1 ms
Wall time: 82.5 ms
4
```