Fix ImpactX Plasma Lens - #222

The class was never exported from `abel/__init__.py` and never
referenced, and had accumulated enough breakage that it could not have
run against a modern ImpactX:

* `TaperedPL(units=0)` - the keyword is `unit`, renamed upstream long
  ago; the call raised `TypeError`.
* `beam2particle_container(beam0, sim)` passed the simulation object
  positionally into the `nom_energy` slot.
* The simulation was built by hand without `particle_shape` or
  `init_grids()`, so it raised `RuntimeError` before tracking.
* The integrated focusing strength was missing its length factor:
  `TaperedPL(unit=0)` takes `k = L * g / (B*rho)`, but the code summed
  to `g * c / E`, i.e. a factor `length` too small.
* The sign hard-coded electrons instead of using `beam.charge_sign()`.
* A `BeamMonitor` was constructed and never used.

`track()` now builds a lattice and delegates to `run_impactx()`, the
same way `InterstagePlasmaLensImpactX` does, which removes the hand-
rolled simulation setup entirely. `num_slices` and `use_apertures` are
exposed; apertures are off by default to match the other `PlasmaLens`
implementations.

Verified against `PlasmaLensNonlinearThick`, which models the same
drift-kick sequence: beam sizes agree to <1e-4 relative and the
offset-induced deflection to <5e-4, across plain / offset / tapered
configurations.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BGAJYtvSuTMamdj5oaTRHz