Use index-friendly CQL for identifiers

[okolawole-ebsco]

Purpose

The previous CQL query for matching instances by identifier was causing significant performance issues when there were many values to filter. The generated query (identifiers="\"identifierTypeId\":\"...\"" AND ...) performed a string-based search on the entire JSONB object without using the index in some cases.

Approach

This refactors the query generation to use CQL relation modifiers, resulting in the format:
identifiers =/@identifierTypeId=... "..."

This is the correct, idiomatic syntax for querying array elements. It allows cql2pgjson to generate an efficient SQL statement that leverages the GIN index, improving query performance by avoiding sequential scans.

From RMB documentation

CQL: @-relation modifiers for array searches

RMB 26 or later supports array searches with relation modifiers, that
are particular suited for structures like:

"property" : [
  {
    "type1" : "value1",
    "type2" : "value2",
    "subfield": "value"
  },
  ...
]

An example of this kind of structure is contributors (property) from
mod-inventory-storage . contributorTypeId is the type of contributor
(type1).

With CQL you can limit searches to property with regular match in
subfield, with type1=value1 with

property =/@type1=value1 value

Observe that the relation modifier is preceded with the @-character to
avoid clash with other CQL relation modifiers.

The type1, type2 and subfield must all be defined in schema.json, because
the JSON schema is not known. And also because relation modifiers are
unfortunately lower-cased by cqljava. To match value1 against the
property contents of type1, full-text match is used.

Multiple relation modifiers with value are ANDed together. So

property =/@type1=value1/@type2=value2 value

will only give a hit if both type1 has value1 AND type2 has value2.

It is also possible to specify relation modifiers without value. This
essentially is a way to override what subfield to search. In this case
the right hand side term is matched. Multiple relation modifiers
are OR'ed together. For example:

property =/@type1 value

And to match any of the sub properties type1, type2, you could use:

property =/@type1/@type2 value

In schema.json two new properties, arraySubfield and arrayModifiers,
specifies the subfield and the list of modifiers respectively.
This can be applied to ginIndex or fullTextIndex.
schema.json example:

{
  "fieldName": "property",
  "tOps": "ADD",
  "caseSensitive": false,
  "removeAccents": true,
  "arraySubfield": "subfield",
  "arrayModifiers": ["type1", "type2"]
}

For the identifiers example we could define things in schema.json with:

{
  "fieldName": "identifiers",
  "tOps": "ADD",
  "arraySubfield": "value",
  "arrayModifiers": ["identifierTypeId"]
}

This will allow you to perform searches, such as:

identifiers = /@identifierTypeId=7e591197-f335-4afb-bc6d-a6d76ca3bace 6316800312

[KaterynaSenchenko]

Let's merge this change in sunflower first to test on PTF #411