add option to skip unused feature pipeline steps

[cportele]

Pull request checklist

• Added/updated unit tests
• Updated documentation
• All checks are passing

Changes introduced by this PR

Closes #1508
Closes #1345

Depends on ldproxy/xtraplatform-spatial#416.

A new feature provider option is introduced to enable skipping unnecessary/unused pipeline steps:

queryProcessing: 
  skipUnusedPipelineSteps: true

Tests with the simple strassenbaukataster data from #1345 and a feature query with a large bounding box that matches 224995 features, varying page sizes, the native CRS as the target CRS, and GeoJSON and FlatGeobuf as output formats:

curl "http://localhost:7080/strassenbaukataster/collections/strassenbaukataster/items?bbox=9.7513%2C53.4112%2C10.2635%2C53.7450&crs=http%3A%2F%2Fwww.opengis.net%2Fdef%2Fcrs%2FEPSG%2F0%2F25832&f=$format&limit=$limit" -o /dev/null

Page size	Format	Time without skipping pipeline steps (ms)	Time with skipping pipeline steps (ms)
1000	GeoJSON	281	209
1000	FlatGeobuf	536	234
10000	GeoJSON	1179	543
10000	FlatGeobuf	1366	393
100000	GeoJSON	8837	3842
100000	FlatGeobuf	11005	2791
224995	GeoJSON	19217	9013
224995	FlatGeobuf	24623	6526

[azahnen]

Deployment of preview was successful: https://ldproxy-ldproxy-pr-1509.surge.sh

[cportele]

@azahnen - Please wait with a review. Currently the changes do not improve performance for queries for generating vector tiles. I will refactor the changes to also support vector tile generation.

[cportele]

@azahnen - This PR is ready for review. Latest changes:

• code moved to xtraplatform-features to support also feature queries for tiles;
• removed the DATE_FORMAT transformation that was added for a property with value type DATE as the SQL response is already in format yyyy-MM-dd;
• split the MAPPING step into two options (mapping of schema and values and value transformations only).

I tested it with a number of API configurations and typically the GeoJSON and FlatGeobuf formats (comparison that the result is the same, excluding timestamps, with and without the option); FlatGeobuf only for the simpler data schemas.

We could also ignore the option for single feature queries or small page sizes as for those cases the overhead to determine the necessary steps is probably similar to potential the savings.