ENH: Improve memory estimates for dense PET and GTM workflows

[mnoergaard]

This PR makes PETPrep’s Nipype scheduling more conservative for memory-heavy PET workflows.

Changes include:

1. Add PET memory estimates for per-frame and reference-generation workloads
2. Apply estimated memory to PET reference nodes
3. Reserve memory for HMC split/smooth/threshold steps
4. Add explicit GTM segmentation memory estimates
5. Scale GTM memory when high-resolution FreeSurfer or submillimeter T1w geometry is detected

This should reduce BrokenProcessPool failures caused by underestimating memory and running too many large nodes concurrently.

[codecov]

Codecov Report

❌ Patch coverage is 97.76119% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 86.38%. Comparing base (0e049fe) to head (ad06ba0).
⚠️ Report is 87 commits behind head on main.

Files with missing lines	Patch %	Lines
petprep/workflows/base.py	82.35%	2 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #328      +/-   ##
==========================================
+ Coverage   86.24%   86.38%   +0.14%     
==========================================
  Files          88       88              
  Lines        9988    10115     +127     
  Branches     1056     1064       +8     
==========================================
+ Hits         8614     8738     +124     
- Misses       1143     1145       +2     
- Partials      231      232       +1     

☔ 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.