Sample each edge type with its own limit, 10 types per request

[kmcginnes]

Description

This is the fix for #2141 on its own, with the existing batching and no new setting or strategy.

• The old Gremlin query grouped every edge of a batch before sampling, and Neptune runs that grouping outside its engine, one item per edge. A graph with three edge types over 19.9M edges ran out of memory in about 30 seconds.
• Each edge type now gets its own limited union() branch, so the 10,000-edge limit applies before anything is read. One limit after hasLabel(A, B, ...) would be shared, and a dominant type would fill it and leave the rest empty.
• Results are grouped by edge type before counting, because DFE couldn't count one projection across several full branches (two took 54s, five timed out; grouping first handled ten in 9s). The anchor is V().limit(1) because inject() isn't native on Neptune, and mid-traversal V() because E() there needs TinkerPop 3.7.
• Batches drop from 100 edge types to 10, so a request reads at most 100,000 edges. 100 types in one request took 116s and left a db.t3.medium refusing even single-type samples for two minutes. Other users of DEFAULT_BATCH_REQUEST_SIZE are unchanged.
• Endpoint labels are folded and every entry is split on ::, so a multi-label vertex keeps every label on Neptune 1.4 (one composite) and 1.3.5 (one entry per label). Before, 1.3.5 silently kept only the first.
• Dedup uses createEdgeConnectionId, so a label containing the old - separator can't collide.

Validation

• The query shape matched the true edge connections on Neptune 1.2.1.0 (15/15), 1.3.5.0 (522/522), 1.4.5.1 (2,033/2,033) and 1.4.7.0 (10,018/10,018), and ran on TinkerPop 3.6.2.
• Four concurrent requests of 10 full edge types each finished in 36s on db.t3.medium instances with and without DFE, and left them answering normally.
• Tests: template shape and escaping, batches of 10, multi-label composites and one-per-entry labels.
• pnpm checks and pnpm test clean on this layer.

Related Issues

• Fixes [3.2.2][Gremlin] Relationship discovery fails with MemoryLimitExceededException for high-cardinality edge labels #2141
• Split out of Ask the database for distinct edge connections instead of scanning edges #2244
• Layer 1 of 4. Stack, bottom first: Sample each edge type with its own limit, 10 types per request #2279 → Show the Schema view without edge connections when their discovery fails #2277 → Ask the database for distinct edge connections instead of scanning edges #2280 → Let a connection choose how edge connections are discovered #2282

Check List

• I confirm that my contribution is made under the terms of the Apache 2.0 license.
• I have verified pnpm checks passes with no errors.
• I have verified pnpm test passes with no failures.
• I have covered new added functionality with unit tests if necessary.
• I have updated documentation if necessary.

[kmcginnes]

Approved