[Lens] Refactor handling of empty values

[flash1293]

Currently, empty values in data table cells are handled by the rename_columns expression function. They will be replaced by the string (empty). This is convenient because all charts will handle it correctly, but there are a few downsides to this approach:

• rename_columns isn't really the expected place for this functionality
• The chart can't differentiate between actually missing values and the (empty) string when filtering (e.g. see [Lens] When filtering for empty field, the filter doesn't filter data correctly #66536 )
• The chart can't reliably style actually empty values in a different way

Possible solutions

Handle this in each renderer

Have a sanitizeValue function in the renderer itself. This is easy to implement, but leads to code redundancy.

Handle this in the field formatter

We have the ability to pass formatting hints along with the data. This formatter could handle empty values in the same way. If this is implemented as part of the existing field formatters, it would also change behaviors in other apps using the infrastructure (Visualize and Discover) - I'm not sure whether this is a good thing, but consistency could be beneficial. Another approach would be a wrapping formatter just for lens handlings this - e.g { id: 'string', params: { transform: 'lowercase' } } becomes { id: 'lens_default', params: { id: 'string', params: { transform: 'lowercase' } } }

[elasticmachine]

Pinging @elastic/kibana-app (Team:KibanaApp)

[stratoula]

Now that TSVB also integrates with the color service, we should ensure that (empty) values will be treated treated the same among Lens, TSVB and XY axis #93749 (comment)