OPT: precompute radar geometry and point weights

[rcjackson]

This pull request further optimizes the optimization engines. The radar geometry and point constraint weights were being needlessly calculated on every function iteration. These are now just computed once and cached for about a 10 percent peformance gain.

Co-written by GPT-5.6-Luna

[codecov]

Codecov Report

❌ Patch coverage is 76.03306% with 29 lines in your changes missing coverage. Please review.
✅ Project coverage is 70.43%. Comparing base (b3c3ec8) to head (50e7230).
⚠️ Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
pydda/retrieval/wind_retrieve.py	63.63%	12 Missing ⚠️
pydda/cost_functions/_cost_functions_numpy.py	75.75%	8 Missing ⚠️
pydda/cost_functions/_cost_functions_tensorflow.py	75.86%	7 Missing ⚠️
pydda/cost_functions/_cost_functions_jax.py	92.30%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #170      +/-   ##
==========================================
- Coverage   70.56%   70.43%   -0.14%     
==========================================
  Files          32       32              
  Lines        5793     5865      +72     
==========================================
+ Hits         4088     4131      +43     
- Misses       1705     1734      +29     

Flag	Coverage Δ
unittests	70.43% <76.03%> (-0.14%)	⬇️

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.