[FEA] Replace post-round cast over decimal round

[sperlingxx]

Is your feature request related to a problem? Please describe.
This is a follow-up issue of #5766 and #5786. Post-round casts are conducted over decimal rounds to fit the output precision and scale: code1 code2 .

Describe the solution you'd like
They are sort of workaround, which can be replaced by cuDF built-in support.

[jlowe]

cudf will short-circuit the cast if the cast ends up being a no-op (e.g.: DECIMAL64 -> DECIMAL64). As such this is relatively minor and would only be a real cast in the case where the pre-cast for round ends up crossing a precision boundary (e.g.: DECIMAL32 -> DECIMAL64).

[jlowe]

Closing this as wont fix for now, this is likely not getting fixed in cudf and performance impact is likely negligible.