Faster is_zero_row for Nemo matrices

[lgoettgens]

the point is that is_zero_row for ZZ should be done via

• get the pointer of the row (no c-call!)
• check if the entries are binary zero (no c-call)
  Currently it uses 2 c-calls per entry hence the generic is fine.
  The same logic would apply also to Z/nZ, small or large and possible Native.GF(p)

Originally posted by @fieker in #1801 (comment)

[fingolfin]

Here are some steps towards that:

1. Replace ccalls in mat_entry_ptr methods with direct pointer computations, e.g. as done in PR Optimize mat_entry_ptr for ZZMatrix #1807
2. Replace ccalls in various test functions with more direct access, e.g. as done in PR Optimize is_unit/is_zero/is_one for ZZRingElem #1808 -- another e.g. also is_zero etc. for QQFieldElem could be im

My two PRs mark the way, but similar techniques should be applied to many other flint wrappers. Perhaps @aaruni96 can work on that after the Begehung (we'll talk about it when he's back in Kaiserslautern)

[fingolfin]

With Nemocas/AbstractAlgebra.jl#1802 added to the mix, I really think we can close this.