Columnar batch indexing follow ups for elastic/logs Rally track.

If columnar batch indexing is enabled in elastic/logs Rally track (simulates workloads from o11y integration) it falls back to sequential execution. This issue is about addressing these issues, so that columnar batch indexing (with elastic/logs rally track) doesn't fall back to sequential execution.

• Support columnar batch indexing for keyword multi-fields (@martijnvg) #158199
• Support columnar batch indexing for match_only_text fields (@martijnvg) #158201
• Support columnar batch indexing for text fields (@seanzatzdev) #158798
• Support columnar batch indexing for wildcard fields (@seanzatzdev) #159263
• Support columnar batch indexing for geo_point fields (fixes source.geo.location in {"lat": ..., "lon": ...} format from GeoIP enrichment) (@seanzatzdev)
• Merge dotted/nested spelling aliases in columnar batch indexing. The rally track's dataset contains field with dots using object notation.
• Skip always-empty-object leaves in columnar batch indexing (fixes "json": {} in k8-application corpus). The rally track datasets contain some empty object fields that are not mapped, triggering sequential path. #159069

These tasks originate from #158130, which investigated the impact on elastic/logs indexing performance if columnar batch indexing is enabled.

Note that if all these tasks are implemented, then only columnar batch indexing's sequential execution path is activated if fields are unmapped and dynamic is true. But the sequential execution will stop as soon as unmapped fields are mapped.