Allow use of string variables in the bucket_script and bucket_selector aggregations

[colings86]

Currently pipeline aggregations always assume they are acting on numerical data. This is fine for most pipeline aggregations since things like moving average and derivative only make sense with numerical data. But for the bucket_script and bucket_selector aggregations it can be useful to also have non numerical data as inputs.

We should remove the explicit check for a numerical value when retrieving the value for just the bucket_script and bucket_selector aggs.

[colings86]

closing in favour of #23874