[magatrons]

Closes #3210

Description Of Changes

Adds metpy.calc.boyden_index, the Boyden Index from Boyden (1963), requested in #3210.
It sits alongside the other sounding indices (total_totals_index, k_index,
sweat_index) and follows the same pattern: same decorators, xarray support through
vertical_dim, and units stripped for the final arithmetic the way sweat_index does
it, as suggested in the issue discussion.

I = (Z700 - Z1000) - T700 - 200

where Z700 - Z1000 is the 1000-700 hPa thickness in decameters and T700 is the
700 hPa temperature in °C. The result is returned as a dimensionless pint.Quantity.

Note on the formula in the issue: the issue writes the index as
Z(700-1000) - T700 - 200 without saying what unit the thickness is in. Both the
AMS Glossary and Boyden (1963) define
the thickness in decameters. With meters the index comes out far outside its usual range
of about 90-100 — for the test sounding here, meters gives 2842.0 instead of the expected
97.0. The tests check against the decameter values, so they would catch that mistake.

Why height is an input: I take geopotential height directly rather than deriving
the thickness from temperature with thickness_hydrostatic. The index is defined in
terms of the observed 1000-700 hPa thickness, and taking height directly keeps the
function simple and leaves out moisture assumptions. If you'd prefer
thickness_hydrostatic (for example, so it can be called with only pressure and
temperature, like the other indices), I'm happy to change it.

Changes:

• src/metpy/calc/thermo.py: new boyden_index(pressure, height, temperature, vertical_dim=0), with docstring and example
• docs/_templates/overrides/metpy.calc.rst: added to the Soundings list
• docs/api/references.rst: added [Boyden1963]
• tests/calc/test_thermo.py: test_boyden_index and test_boyden_index_xarray

Testing:

• The new tests use expected values worked by hand, with sounding levels placed exactly
  on 1000 and 700 hPa so interpolation doesn't affect them. The xarray test uses
  temperatures in kelvin, so it also exercises the conversion to °C.
• Full suite locally: 1626 passed, 2 xfailed, 8 failed. The 8 failures are all in
  tests/remote/test_aws.py and fail identically on unmodified main. They come from my
  local environment, not this change: a non-default AWS region in my config makes the
  requests miss the recorded cassettes, and test_goes_single makes a live download that
  fails on my machine. With my AWS config disabled, only test_goes_single fails.
• ruff check, flake8, doc8 docs and codespell all pass, and the doctest passes.

Checklist

• Closes metpy.calc.boyden_index #3210
• Tests added
• Fully documented

[CLAassistant]

All committers have signed the CLA.