doc: fix IrreducibleModules description to say 'at most' dim

[mvanhorn]

Closes #6004.

Problem

The GAPDoc description for IrreducibleModules at lib/grpreps.gd:49-53 claimed the second return entry was:

a list of all irreducible modules of G over the field F in dimension dim

The actual behaviour is to return all irreducible constituents whose dimension is at most dim, same as the sibling operation AbsolutelyIrreducibleModules.

Evidence

Two independent confirmations in the same file / codebase:

1. Sibling operation documents it correctly. AbsolutelyIrreducibleModules at lib/grpreps.gd:18-23 already says "can be realized over the finite field F and have dimension at most dim". Both operations share the same (G, F, dim) signature and are implemented side by side; the description should match.

2. InstallOtherMethod comment confirms 0 means "no bound". lib/grpreps.gi:314 says:

   InstallOtherMethod(IrreducibleModules,"Supply no dimension limit",
     [IsGroup, IsField, ...], 0,
   function( G, F ) return IrreducibleModules(G,F,0); end);
   

   Only makes sense if dim=0 lifts the upper bound, not if it selects modules of dimension exactly zero.

Change

Two-line diff to lib/grpreps.gd:

-##  <A>G</A> over the field <A>F</A> in dimension <A>dim</A>, given as MeatAxe modules
+##  <A>G</A> over the field <A>F</A> of dimension at most <A>dim</A>,
+##  given as MeatAxe modules
 ##  (see&nbsp;<Ref Func="GModuleByMats" Label="for generators and a field"/>).
+##  Pass <C>0</C> for <A>dim</A> to impose no dimension bound.

The Pass 0 to impose no dimension bound sentence is a small addition that surfaces existing behaviour — previously only discoverable from the InstallOtherMethod comment — directly in the reference manual.

Docs only. No behaviour change.

[fingolfin]

@mvanhorn hi there, thanks for your contribution.

As per our AI policy,

Any use of AI tools for preparing code, documentation, tests, commit messages, pull requests, issue comments, or reviews for this repository must be disclosed

[mvanhorn]

@fingolfin Apologies for missing the disclosure up front. This PR was prepared with Claude Code (Anthropic's AI coding assistant): it helped me navigate the grpreps sources, confirm the "at most" semantics via the sibling operation and the InstallOtherMethod comment, and draft the two-line doc change. I reviewed the diff and PR body before opening.

Happy to add a disclosure line to the PR description, or to close if you would prefer a non-AI-assisted submission.

[fingolfin]

It's fine to use AI assistance, just disclose it (e.g. via a Co-authored-by line in the commit message or so).

[mvanhorn]

@fingolfin Done. Amended commit b5ec59c with Co-authored-by: Claude (Anthropic AI) to disclose AI assistance per your preference. Diff unchanged.

[mvanhorn]

Thanks for landing the doc fix, @fingolfin.