feat!: Add multi-column support to AggFormula

[lbooker42]

Python examples:

from deephaven import empty_table
from deephaven import agg

source = empty_table(20).update(
    ["X = i", "Y = 2 * i", "Z = 3 * i", "Letter = (X % 2 == 0) ? `A` : `B`"]
)

result = source.agg_by([
    agg.formula(formula="out_a=sqrt(5.0)"),
    agg.formula(formula="out_b=min(X)"),
    agg.formula(formula="out_c=min(X) + max(Y)"),
    agg.formula(formula="out_d=sum(X + Y + Z)"),
], by=["Letter"])

Groovy examples:

source = emptyTable(20).update("X = i", "Y = 2 * i", "Z = 3 * i", "Letter = (X % 2 == 0) ? `A` : `B`")

result = source.aggBy([
    AggFormula("out_a=sqrt(5.0)"),
    AggFormula("out_b=min(X)"),
    AggFormula("out_c=min(X) + max(Y)"),
    AggFormula("out_d=sum(X + Y + Z)"),
], "Letter")

[chipkent]

Supporting cols as a List[str] seems to suggest that we might want to support a list of formulas.

[rcaudy]

I'm not sure

formula(['output_col1=(input_col1 + input_col2) * input_col3', 'output_col2=(input_col21 + input_col22) * input_col23'])

is better than

[formula('output_col1=(input_col1 + input_col2) * input_col3'), formula('output_col2=(input_col21 + input_col22) * input_col23')]

[chipkent]

Our other methods do the second. I'm pointing out the inconsistency.

[lbooker42]

There is a technical reason why we can't do that immediately for Java (already using String... var args to support Pair args). We could do it for Python but then we'd have a mismatch.

When/if we completely sunset the parameterized version, we can implement this without any conflict.

[devinrsmith]

It's worthwhile bringing up the gRPC structuring layer given work that Charles want to do there too...

[rcaudy]

.

[rcaudy]

.