Use weighted cHPI coils during position fitting

[larsoner]

Closes #11330

This turned out to way simpler than I realized:

1. Use a linear ramp from gof_limit to 1.
2. Combine this with the dist_limit coil goodness estimate
3. Pass these weights to our quat fitting function, which already supports weights

Implementation drafted with Opus 4.8 but I edited and vouch for all code.

Adapting the minimal example from the linked issue (data available there):

Minimial code

import mne
raw_file = 'te01a1.fif'
raw = mne.io.read_raw_fif(raw_file)
chpi_amps = mne.chpi.compute_chpi_amplitudes(raw, t_step_min=0.25, t_window=0.5)
chpi_locs = mne.chpi.compute_chpi_locs(raw.info, chpi_amps)
head_pos = mne.chpi.compute_head_pos(raw.info, chpi_locs, weighted=False)
mne.viz.plot_head_positions(head_pos)
head_pos = mne.chpi.compute_head_pos(raw.info, chpi_locs, weighted=True)
mne.viz.plot_head_positions(head_pos)

weighted=False	weighted=True