This is because I noticed the assembly format was printing <10> for
polynomials instead of the fully qualified type name. After this commit
it will print the whole type

See https://mlir.llvm.org/docs/DefiningDialects/Operations/#declarative-assembly-format
for more details.

This removes the flexibility of having mixed poly + tensor ops for the
binary operations, but demonstrates how the type inference engine
enables a more succinct textual IR.

If you were to simplify the assembly format without doing this, you'd
get a compile-time error complaining that it can't infer the type of the
operands or argument.