Let a plot be drawn in a unit of the caller's choosing

[roytsmart]

A system is free to describe its parts in whichever units suit them, and a
detector measured in microns beside optics measured in millimeters is quite
normal. ESIS is built that way: its sensor comes from msfc-ccd, where a pixel
pitch is naturally microns, while the optics are millimeters.

Nothing reconciles those units on the way to matplotlib.
astropy.visualization.quantity_support does it on a 2D axes, but not on a
3D one, so a surface given in microns is drawn a thousand times too large:

d = esis.flights.f1.optics.design(num_distribution=0)
ax = plt.figure().add_subplot(111, projection="3d")
d.system.plot(ax=ax, components=("y", "z", "x"), plot_rays=False)
ax.get_zlim()   # (-121491.1, 121491.1)

That is microns being read as millimeters. The instrument collapses to a speck
and the detectors land outside the frame entirely, with nothing to say so. It
is a quiet failure: the geometry is correct, only the scale it is drawn at is
wrong, so a layout figure simply comes out missing its detectors.

The change

plot gains a unit, and converts every length to it before drawing: the
surfaces, both of their apertures, and the rays.

d.system.plot(ax=ax, components=("y", "z", "x"), plot_rays=False, unit=u.mm)
ax.get_zlim()   # (-135.8, 135.8)

The default is unchanged, so nothing that works today moves.

This mirrors _write_to_dxf, which sits directly below plot in each of these
classes and has always taken a unit and converted. The plotting path was the
one that did not.

Why not fix the sensor

The other option was to have the sensor describe itself in millimeters, as the
older kgpy model did, where the detector's clear width was computed as
(width_pixels * shape_pixels).to(u.mm). That would fix ESIS and no one else,
and it puts the burden on how a part is specified rather than on how it is
drawn. Specifying a pixel pitch in microns should stay comfortable.

Testing

test_plot_unit builds the same aperture twice, once in millimeters and once
in microns, and checks the magnitude handed to matplotlib. Without a unit the
two differ by a factor of a thousand; with one they agree; and asking for
microns converts rather than relabels. I checked it fails without the change.

The existing suites pass: 1641 in apertures, 451 in systems/_sequential.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.58%. Comparing base (beadb92) to head (3518101).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #201   +/-   ##
=======================================
  Coverage   99.58%   99.58%           
=======================================
  Files         118      118           
  Lines        6933     6963   +30     
=======================================
+ Hits         6904     6934   +30     
  Misses         29       29           

Flag	Coverage Δ
unittests	99.58% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.