Avoid repeated outer products in free-bound emission calculation

[wtbarnes]

Fixes #318

This refactors the Ion.free_bound calculation a bit to avoid repeated np.outer calls that produce arrays of size n_temperature by n_wavelength.

Doing some informal benchmarking, without this PR, for a temperature array of length 5000 and a wavelength array of 1000,

============================================================================ test session starts ============================================================================
platform darwin -- Python 3.12.3, pytest-8.2.1, pluggy-1.5.0
rootdir: /Users/wtbarnes/Documents/codes/fiasco
configfile: pyproject.toml
plugins: astropy-0.11.0, cov-5.0.0, remotedata-0.4.1, anyio-4.3.0, filter-subpackage-0.2.0, doctestplus-1.2.1, astropy-header-0.2.2, hypothesis-6.102.6, asdf-3.2.0, arraydiff-0.6.1, mock-3.14.0
collected 1 item

fiasco/tests/test_performance.py .                                                                                                                                    [100%]

============================================================================= slowest durations =============================================================================
36.18s call     fiasco/tests/test_performance.py::test_fb_speeds
1.05s setup    fiasco/tests/test_performance.py::test_fb_speeds

(1 durations < 0.005s hidden.  Use -vv to show these durations.)

and with this PR,

============================================================================ test session starts ============================================================================
platform darwin -- Python 3.12.3, pytest-8.2.1, pluggy-1.5.0
rootdir: /Users/wtbarnes/Documents/codes/fiasco
configfile: pyproject.toml
plugins: astropy-0.11.0, cov-5.0.0, remotedata-0.4.1, anyio-4.3.0, filter-subpackage-0.2.0, doctestplus-1.2.1, astropy-header-0.2.2, hypothesis-6.102.6, asdf-3.2.0, arraydiff-0.6.1, mock-3.14.0
collected 1 item

fiasco/tests/test_performance.py .                                                                                                                                    [100%]

============================================================================= slowest durations =============================================================================
20.20s call     fiasco/tests/test_performance.py::test_fb_speeds
1.08s setup    fiasco/tests/test_performance.py::test_fb_speeds

(1 durations < 0.005s hidden.  Use -vv to show these durations.)
============================================================================ 1 passed in 21.36s =============================================================================

For smaller arrays, $O(10^2)$ entries, the gains are more modest, closer to a 1 s difference.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 90.47%. Comparing base (a896483) to head (2caf81f).
Report is 17 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #319   +/-   ##
=======================================
  Coverage   90.46%   90.47%           
=======================================
  Files          40       40           
  Lines        3367     3369    +2     
=======================================
+ Hits         3046     3048    +2     
  Misses        321      321           

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