[stertooy]

Extracted from #174.

This caches small negative powers of elements of the currently computed igs.

I added some (small?) improvement of my own:

• Previously, when computing ind[d]^-q, this only tried to either get this exact value from the cache, or get ind[d]^-1 from the cache and then compute (ind[d]^-1)^q. I made it so this also tries to find some ind[d]^-s from the cache and then compute (ind[d]^-s)^(q/s) (where of course s divides q).
• This function was not used in the "add powers and commutators" parts and in the "ReduceExpo" functions... Meaning it often would not draw a known value from the cache we created.
• Use Unbind instead of setting things to [].

Original PR disclaimer:
Assisted-by: Claude Code (Fable 5.1)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 71.61%. Comparing base (7e84dbe) to head (b77a4ca).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #277      +/-   ##
==========================================
+ Coverage   71.51%   71.61%   +0.10%     
==========================================
  Files          88       88              
  Lines       13029    13076      +47     
==========================================
+ Hits         9318     9365      +47     
  Misses       3711     3711              

Files with missing lines	Coverage Δ
gap/basic/pcppara.gi	93.80% <100.00%> (+0.46%)	⬆️
gap/basic/pcppcps.gi	84.95% <100.00%> (+0.76%)	⬆️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[stertooy]

Oh, also, if #284 gets merged first, this PR would benefit from a small change: the "reuse a cached power ... " strategy should be avoided if g is a power of a generator, because such powers are computed much faster due to #284.