Improve performance in parallel case

[timmens]

In this PR, we improve the performance in the parallel case.

More specifically, we

• Consider sampled line search points when drawing a speculative sample
• Set the radius from which we draw the speculative sample to 75% of the trust-region radius
• Add the filter option n_max_factor that allows us to compute a target size for the filtering step dependent on the regular target size

[codecov]

Codecov Report

Attention: 1 lines in your changes are missing coverage. Please review.

Comparison is base (a94c8b5) 78.48% compared to head (cb35dec) 78.64%.

Files	Patch %	Lines
src/tranquilo/acceptance_decision.py	75.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main       #9      +/-   ##
==========================================
+ Coverage   78.48%   78.64%   +0.15%     
==========================================
  Files          68       68              
  Lines        4923     4944      +21     
==========================================
+ Hits         3864     3888      +24     
+ Misses       1059     1056       -3     

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