core: (affine) Add Affine.binary

[watermelonwolverine]

Cherry picked from #3650. Prerequisite for implementing vector.transfer_read and vector.transfer_write ops

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 90.23%. Comparing base (e209b0c) to head (06a5eba).
Report is 4 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #4006      +/-   ##
==========================================
- Coverage   90.23%   90.23%   -0.01%     
==========================================
  Files         458      458              
  Lines       58409    58455      +46     
  Branches     5691     5697       +6     
==========================================
+ Hits        52707    52747      +40     
- Misses       4320     4323       +3     
- Partials     1382     1385       +3     

☔ View full report in Codecov by Sentry.
📢 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.

[watermelonwolverine]

While implementing this I realized that when composing two affine maps the expressions are not automatically simplified/constant folded.
So, a resulting AffineBinaryOpExpr where both lhs and rhs are AffineConstantExpr is not simplified to a AffineConstantExpr.
That's why I added the AffineExpr.as_constant function. Is this by design or should the constant folding happen while composing?

[superlopuh]

What does MLIR do here? I have a suspicion that they constant fold when creating the affine expr, it would be worth doing the same in xDSL. I can't find these helpers in the AffineExpr and AffineMap header files.

[watermelonwolverine]

MLIR simplifies while composing. This is the call stack, the simplification happens at the last call:

AffineExpr AffineExpr::compose(AffineMap map)
AffineExpr AffineExpr::replaceDimsAndSymbols(ArrayRef<AffineExpr> dimReplacements,ArrayRef<AffineExpr> symReplacements)
AffineExpr mlir::getAffineBinaryOpExpr(AffineExprKind kind, AffineExpr lhs, AffineExpr rhs)
AffineExpr AffineExpr::operator+(AffineExpr other)
static AffineExpr simplifyAdd(AffineExpr lhs, AffineExpr rhs)

[watermelonwolverine]

Added AffineExpr.binary which mimics AffineExpr mlir::getAffineBinaryOpExpr from AffineExpr.cpp.
Now it should be pretty close to how MLIR is doing it