Reject zero substitutions in Laurent denominators

[pzinn]

I forget if I ever submitted an issue:

   R=QQ[x,Inverses=>true,MonomialOrder=>Lex]
   f=matrix{{x^-1,1/2}}
   sub(f,x=>0)

will happily produce matrix{{0,1/2}} instead of throwing an error. Fixed with test.

AI Disclosure

codex helped me diagnose this.

[pzinn]

Interestingly, this PR unearthed a bug in an example of InvariantRing:
https://www.unimelb-macaulay2.cloud.edu.au/usr/share/doc/Macaulay2/InvariantRing/html/_equivariant__Hilbert__Series.html
Note that last substitution at z_0, z_1 = 0.
@LuigiFerraro88 @galettof @fragandi @whuang259 @mnmastro @xlni

[mahrud]

This seems reasonable to me, but I fear the chance of conflicting with the two engine refactoring PRs is significant at the moment, so it would make sense to wait until those are merged first, then apply this small fix.

[galettof]

Of course this makes sense mathematically. However, it was a convenient "feature" to able to use sub to extract the part of the polynomial that is constant in the z variables. I have some code that works but it is not particularly elegant.

R = QQ[x_1..x_3]
W = matrix{{-1,0,1},{0,-1,1}}
T = diagonalAction(W, R)
E = equivariantHilbertSeries(T, Order => 7)
(Z,phi) = flattenRing ring E;
(product toList coefficients(phi E,Monomials=>apply(7,i->Z_0^i)))_(0,0)

Surely, there must be a better, less ad-hoc way to achieve something like this. Any suggestions?