implement vectors mod 2 as M4RI matrices with one row

[malb]

In order to resolve the embarrassing situation at #3684, we need faster vectors mod 2.

[malb]

The speed-up provided by this patch is considerable (but not as much as one would hope for). Here's the example from #3684:

Before

sage: A = random_matrix(GF(2),1000,2000)
sage: %time K = A.right_kernel()
CPU times: user 11.24 s, sys: 0.02 s, total: 11.25 s
Wall time: 11.42 s

After

sage: A = random_matrix(GF(2),1000,2000)
sage: %time K = A.right_kernel()
CPU times: user 0.15 s, sys: 0.00 s, total: 0.15 s
Wall time: 0.16 s

Speed-up: 11.25/0.15 = 75

---

Another example

Before

sage: VS = VectorSpace(GF(2),10^4)
sage: e = VS.random_element()
sage: f = VS.random_element()
sage: %timeit e+f
10000 loops, best of 3: 89.8 µs per loop
sage: %timeit e*f
10000 loops, best of 3: 146 µs per loop

After

sage: VS = VectorSpace(GF(2),10^4)
sage: e = VS.random_element()
sage: f = VS.random_element()
sage: %timeit e+f
1000000 loops, best of 3: 1.08 µs per loop
sage: %timeit e*f
100000 loops, best of 3: 2.47 µs per loop

Speed-ups: 83 and 59.

---

The attached patch might indeed depend on #3684.

[mhansen]

Looks good to me.