[Bridges] fix bug in HermitianToSymmetricPSDBridge

[odow]

Closes #2170

[odow]

@blegat, I didn't really understand your indexing logic, so I just started from scratch, working my way through the result vector in a single pass.

inverse_map_function is correct, but you should take a look at adjoint_map_function. I don't really understand what it's doing or if it's correct.

[odow]

cc @araujoms

You can try this branch with:

] add MathOptInterface#od/fix-hermitian-bridge

[araujoms]

Just tested it, works like a charm.

[blegat]

Arf, I messed up the adjoint of the imaginary part, that's the issue when coding without a pen and paper. It would be quite unreadable to maintain the index as it is moving horizontally and not vertically (because of the transpose) so I switched to using trimap