Optimize summary stats mode: cache XtX*beta to eliminate 3x redundant…

[gaow]

… O(P^2) computation

In summary statistics mode, XtX %*% beta was computed 3 separate times per iteration per outcome: in residual update, profile loglikelihood, and correlation update (get_correlation). This commit caches the product once after the beta update and reuses it, reducing the dominant O(P^2) cost by 3x.

Also precomputes per-outcome constants (scaling_factor, beta_scaling) during model initialization to avoid repeated conditional evaluation per iteration.

Benchmark results (micro-benchmark on XtX %*% beta):
P=1000, L=2, M=100: 0.43s -> 0.14s (3x speedup)
P=2000, L=5, M=200: 9.75s -> 3.25s (3x speedup)
P=5000, L=10, M=500: 324s -> 108s (3x speedup, saves 216s)

[codecov]

Codecov Report

❌ Patch coverage is 94.44444% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 84.49%. Comparing base (2b5e1e1) to head (dac0997).
⚠️ Report is 9 commits behind head on main.

Files with missing lines	Patch %	Lines
R/colocboost_update.R	90.90%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #132      +/-   ##
==========================================
+ Coverage   84.05%   84.49%   +0.44%     
==========================================
  Files          14       14              
  Lines        4828     4837       +9     
==========================================
+ Hits         4058     4087      +29     
+ Misses        770      750      -20     

☔ View full report in Codecov by Sentry.
📢 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.