RA wrapping in `interpolate_ephem`

Problem

interpolate_ephem interpolates ephem_ra directly. A track crossing RA of 0 is stored wrapped into [0, 2pi), so the samples contain a 2pi jump that the interpolant reads as real motion, and it sends the target the "long way" around the sky. Found while implementing moving near-field support (#1710). Deliberately not fixed in that PR series because it is a pre-existing bug that affects ephem phase centers and deserves its own PR.

Because the interpolatant is a cubic spline for 4 or more samples, this is not a local error at the crossing; the discontinuity rings across the entire track.

Evidence

A target drifting steadily at 0.05 rad/sample through RA = 0, with 5 samples:

stored ephem_ra   : [6.2    6.25   0.0168 0.0668 0.1168]
continuous form   : [6.2    6.25   6.3    6.35   6.4   ]

Interpolated at every midpoint between samples, compared as angles on the circle so the wrap itself is not counted as error:

All four midpoints are wrong, including even the two before the crossing.

Fix

Interpolate an unwrapped copy of ephem_ra, and fold the result back into [0, 2pi) range on return.

Add tests of ephem and near_field tracks crossing RA of 0 are interpolated correctly, assert non-wrapping track results remain unchanged, and that interpolated returned RA is in [0, 2pi) range.

Start in src/pyuvdata/utils/phasing.py at interpolate_ephem and trace how ephem_ra is interpolated and returned. Add coverage for ephem and near_field tracks crossing RA 0, plus a non-wrapping track and the returned [0, 2pi) range. Done means midpoint interpolation follows the continuous track without changing non-wrapping results.