Implement INV_TRANSAD for the GPU branch.

[l90lpa]

This draft PR implements INV_TRANSAD for the GPU branch. I've reused some existing sudroutines from the direct code, TRGTOL, FTDIR and LEDIR, as on inspection they appear to be the same as the adjoints of TRLTOG, FTINV and LEINV. For TRMTOL routines I've added an adjoint as the pack/unpack of TRLTOM handle scaling and truncation as well as transposition. Also, I've extended the existing trani invtrans_adjoint test to check the vor/div fields, and add a trans test of INV_TRANSAD where I've reused some parts of @samhatfield work in #214. One thing I haven't done is fix-up the GSTATS codes as I'm not exactly sure how the IDs are assigned.

[FussyDuck]

All committers have signed the CLA.

[samhatfield]

Thanks @l90lpa - we'll gradually take a look at this. For now I just added some comments to INV_TRANSAD which I spotted when diff'ing against the CPU version.

[wdeconinck]

Thanks @l90lpa for taking this on!!!

[l90lpa]

Thanks for taking a first pass @samhatfield, I'll make those fixes next week (and any others as needed). @wdeconinck it's no problem!

[samhatfield]

Thanks for taking a first pass @samhatfield, I'll make those fixes next week (and any others as needed). @wdeconinck it's no problem!

I'm just working on a pull request directly to your branch which includes the changes above, so no effort required on your part @l90lpa (other than to merge that PR once it's ready!).