Consolidate statistics aggregation

### Is your feature request related to a problem or challenge?

There are at least three places in DataFusion where multiple `Statistics` objects are aggregated together, and they do so inconsistently:

1. `get_statistics_with_limit`: https://github.com/apache/arrow-datafusion/blob/e54894c39202815b14d9e7eae58f64d3a269c165/datafusion/core/src/datasource/statistics.rs#L34-L33
2 . Parquet::infer_stats: https://github.com/apache/arrow-datafusion/blob/a892300a5a56c97b5b4ddc9aa4a421aaf412d0fe/datafusion/core/src/datasource/file_format/parquet.rs#L503-L581
3. Union::statistics: https://github.com/apache/arrow-datafusion/blob/c2e768052c43e4bab6705ee76befc19de383c2cb/datafusion/physical-plan/src/union.rs#L612-L611

(and we actually have another version of this in IOx)

### Describe the solution you'd like

I would like to consolidate the three implementations into a `StatisticsAggregator` that knows how to aggregate multiple `Statistics` objects that is both documented and well tested.

### Describe alternatives you've considered

_No response_

### Additional context

_No response_

Compare the aggregation logic in datafusion/core/src/datasource/statistics.rs, datafusion/core/src/datasource/file_format/parquet.rs, and datafusion/physical-plan/src/union.rs. Start from get_statistics_with_limit, Parquet::infer_stats, and Union::statistics; done means the shared StatisticsAggregator is documented, the three implementations use it, and its behavior is covered by tests.

Written by the indexing model from the issue text.