Fix InvariantQuadraticForm for Omega(-1, 2*d, 2^n)

[fingolfin]

Also add proper tests to check that the quadratic forms are indeed correct
(the existing tests were too weak).

Fixes #4323

Some background: We didn't even store the InvariantQuadraticForm for these groups up to GAP 4.9; it was only added in 4.10, via my PR #2577. In there, I also added tests to verify the quadratic form, but those tests were insufficient (rather obviously so, in retrospect sigh). This adds a test similar to what Thomas used to highlight the issue in his bug report.

[ThomasBreuer]

Thanks.
Is there anything one can say why the matrix of the quadratic form in OmegaMinus can look like this? (The structure of the generators yields that one has to work only for d = 4.)
In CheckQuadraticForm, checking the relation between the matrices of the bilinear and the quadratic form still makes sense, it is just not sufficient in even characteristic.

[hulpke]

Thanks.
Is there anything one can say why the matrix of the quadratic form in OmegaMinus can look like this? (The structure of the generators yields that one has to work only for d = 4.)

I believe the generators are taken from `Rylands, Taylor: Matrix generators for the Orthogonal Group, JSC 25 (1998), 351--360, and p. 359 (about the middle) gives the form for Omega-

[fingolfin]

@ThomasBreuer I restored the check for the relation between the matrices of the bilinear and the quadratic form -- I did not mean to disable it in the first place, thanks for spotting this!

[ThomasBreuer]

@hulpke Thanks for the hint. Yes, the Rylands/Taylor paper is mentioned in the GAP manual, and it describes the form.
Note that the paper does not distinguish odd and even characteristic. Following the paper, one could simplify the proposed code by setting x[m+1,d-m+1] := -xi;.

[fingolfin]

@ThomasBreuer ah that's of course much nicer. Pushed it.