eja: drop custom gram_schmidt() routine that isn't noticeably faster.
Using Gram-Schmidt in SageMath is a bit of a pain, since you have to
put your vector space elements into a matrix, which means you have to
convert them to vectors first. And then you wind up transposing the
matrix twice, and then you have to convert everything back to vectors
and then back to algebra elements, dropping the zero vectors/elements
in the process. Sounds slow, right?
Well, testing shows that jumping through all those hoops and using the
default implementation isn't much slower than the custom routine I
wrote. So, for now, let's delete it.
Beware that this breaks the construction of subalgebras when the field
is not exact because SageMath won't orthonormalize over an inexact
field.