Add a native Fortran .param file reader

[neuromechanist]

Background

Reviewer feedback on the JOSS paper (Arnaud Delorme) asked whether pyAMICA and the
reference Fortran binary can be driven from the same configuration file. Right now
the answer is: partially.

AMICA.from_params_file (pyAMICA/amica.py) reads a JSON file
(sample_data/sample_params.json), whose keys mostly reuse Fortran input.param
names (lrate, do_newton, rho0, block_size, max_iter, num_models, ...),
but:

• not all keys match one-to-one -- e.g. num_mix (JSON) vs num_mix_comps
  (Fortran input.param)
• pyAMICA has no reader for the literal Fortran .param text format at all, so a
  Fortran input.param file cannot currently be handed to pyAMICA directly

Proposal (revised: translator, not a second parser)

We already have a working JSON-parameter code path (from_params_file ->
constructor kwargs). Rather than teaching that path to also parse Fortran syntax,
add a small translator: parse the plain-text key value lines of a Fortran
input.param file into a dict, rename the handful of mismatched keys
(num_mix_comps -> num_mix, etc. -- full mapping TBD by diffing
sample_params.json against input.param), and hand the result to the exact same
construction path from_params_file already uses. No new parameter-handling logic,
just an input-format adapter in front of the existing one.

Document the key-mapping table alongside docs/guides/validation.md#parameter-files.

Why

So a single parameter file can drive both the Fortran binary and pyAMICA for a
parity run, instead of maintaining a hand-translated JSON copy.

Not required for the current JOSS submission; tracked here as anticipated follow-up
work per reviewer feedback.