Use is_zero_approx and fix spelling in CameraMatrix invert

[aaronfranke]

This PR fixes some problems noticed in #58492 (comment)

• Fix spelling of determinant (was "determinat").
• Fix Math::absd being used on real_t when it should be Math::abs.
• Use is_zero_approx() instead of absd() < 1e-7 since 1e-7 is so small of an epsilon to the point of not being very useful (float precision around 1 is 2^-23 or about 1.192e-7). It's possible that 1e-7 works for this case, but if that is true, then why it's so small needs to be explained in a comment. This part needs review from @reduz.

[lawnjelly]

Looks good.

I'm not absolutely sure on the epsilon either though so could do with another pair of eyes, I suspect that this is to cope with the case where the matrix is very small scale, or the vectors are colinear in which case I don't think you'll get anything sensible out of it (?).

I suspect anything slightly above zero in the resolution will do the job, as you'll probably see nothing on screen with such a CameraMatrix. Willing to be corrected by someone more familiar with the maths though. 😄

[lawnjelly]

Note In rocket chat reduz also thought this should probably just be CMP_EPSILON, which is what is_zero_approx() does under the hood. So should be good barring the dead code that akien asked about. 👍

[akien-mga]

Thanks!