[inamdarmihir]

engine/clients/elasticsearch/configure.py refuses any dataset with vector_size > 2048 by raising IncompatibilityError. That limit was the Elasticsearch 8.10 dense_vector maximum (the comment links the 8.10 docs). The benchmark's server image is now elasticsearch:9.5.1, where dense_vector accepts up to 4096 dimensions, so 3072/4096-dim datasets are skipped for no reason. The blocker noted on #113 ("need to update the ES version") no longer applies.

This replaces the literal with a MAX_DENSE_VECTOR_DIMS = 4096 constant and updates the docs link.

Checked against Elastic's source: DenseVectorFieldMapper.MAX_DIMS_COUNT is 2048 at tag 8.10 and 4096 at 8.11, 8.12, 9.3.0 and main; the docs say it "can't exceed 4096". I did not check tag 9.5.1 itself. The OpenSearch client has its own, different limit (1024) and is untouched.

Tests (tests/engine/clients/elasticsearch/test_elasticsearch_configure.py, stubbed ES client, no server needed): 128, 2048, 2049 and 4096 dims are accepted; 4097 and 8192 raise IncompatibilityError without calling indices.create; the existing Distance.DOT rejection still works.

pytest tests/engine/clients/elasticsearch   # 8 passed

Fixes #113