[FEA] - AST Support for fixed_point literals.

[breta-nv]

I wish I could use fixed_point literals in AST.

Our team utilizes compute_column, and we'd like to use fixed_point literals in the ast::expression. Without fixed_point, we have to rely on floating-point, which leads to incorrect results due to the usual precision and rounding errors.

Without AST support, we have to rely on a slower fallback path.

[JayjeetAtGithub]

In #16088, I added support for fixed_point_scalar literals in AST here, which allows me to now pass fixed_point_scalar in an cudf::ast::operation. Although, the code compiles now, I am getting some weird behaviour when I am applying such an expression to cudf::compute_column.

So, basically, I have 2 filter expressions, each of them has a fixed_point_scalar. In one of them the value is 0.05 and in the other the value is 24. I am representing both these values as a fixed_point_scalar with a defined scale of -2 (same as my columns scale). The filter with 24 works correctly but the one with 0.05 does not. I am still trying to figure it out. But, I would like to know if anything is weird, or wrong that I am doing in this whole setup. Thanks a lot.