Out of bounds reads

[mlourakis]

Hi,

I needed a bare bones QR and your code came in handy. Going though it, I realized that it had an out of bounds error in partialdot_product() which I have fixed in this PR. Interestingly, this error does not seem to affect the computed factorization.

If called with length=10 and index=1, the original code will access vector elements at indices
1 2 3 4 5 6 7 8 9 10 (all in the 2nd loop)
Note that the last one (10) is out of bounds.

Similarly, for a call with 10 2, the code accesses
2 3 4 5 6 7 8 9 10 11
where the last two indices are again out of bounds.

With the new code, for (10, 1) the indices accessed are
1 2 3 4 (1st loop)
5 6 7 8 9 (2nd loop)
and are all within bounds.

For the call (2, 10), the new indices are
2 3 4 (1st loop)
5 6 7 8 9 (2nd loop)