Fix the BORGLPT transfer function

[maho3]

Currently, there is a power discrepancy between outputs of BORGLPT and pmwd's internal 2LPT.

Everything seems to be in the right configuration, but at about 30% difference in power. This power differential seems to be dependent on cosmology (linearly anticorrelated with Omega_m).

The likely culprit is that, though the transfer function for both codes is Eisenstein-Hu, pmwd fits A_s with a boltzmann code, while BORG seems to use some sort of crude emulator. This results in a different initial scaling to the power spectrum.

Solutions:

• BORG includes a CLASS transfer function which fits A_s, but it is not available in the public version.
• jax_lpt will soon have a public version which does not depend on BORG and can provide BORG-CLASS-like A_s fitting transfer functions
• We can use pmwd's transfer function to seed BORGLPT's ICs

As a temporary solution, I think we should pursue the latter. However, when jax_lpt becomes independent of BORG we should consider using that one to make everything self-consistent

[maho3]

Somehow, in the new versions of pmwd, this isn't a problem.

Either they fixed pmwd to match the BORG LPT, or pmwd's PM code can automatically take its own LPT and produce the right z=0 fields. I tested this exhaustively across many cosmologies and both BORG-LPT and PMWD were consistent with Quijote.

In any case, this isn't a problem anymore.