Upgrades Elasticsearch integration to support v8+

[jerluc]

Overview

The main change included here is around dropping support for Elasticsearch v6 and adding new support for Elasticsearch v8 (there is no plan at the moment to support multiple simultaneous versions). This is meant to both support major infrastructure changes @ Station A while also addressing some historical feedback we've gotten here (e.g. #51 and #56).

What's changed?

In order to support Elasticsearch v8, quite a few things were needed:

• Per Upgrade Elasticsearch client libraries #56, the olivere/elastic library we were originally using is no longer maintained; instead, this implementation uses the official Elasticsearch Go client
• The new implementation now makes use of Elasticsearch's native vector tile search functionality; this has some notable implications compared to the original scroll-based implementation (though I think these are reasonable trade-offs for now):
  • Tiles are now capped at a maximum of 10,000 features
  • Feature properties are flattened by Elasticsearch before tilenol remaps them
  • Some property values (e.g. arrays) are not supported well
  • Vector tiles are actually re-encoded in order to remap fields and rename layers appropriately--this may become a source of performance issues in the future
• Breaking changes to the YAML configuration for Elasticsearch layer sources:
  • host + port are now combined into the hosts array
  • New username and password options for clusters secured via HTTP basic auth
  • Removed the sniff option

Other notables

• A couple of changes to the examples/index.html demo map page
  • Feature popup window now automatically overflows when properties run long
  • Mapbox access token must be set via the mapbox_access_token query parameter
  • All other query parameters get forwarded to tilenol so you can try out additional tilenol parameters such as q=<QUERY-STRING-QUERY> or s=<EXTRA-FIELD-MAPPING>
• Updated the Docker images to Go v1.22 and Alpine v3.20

[schimmy]

LGTM, just one nit