Remove task_execution_id column from task_instances table

[mcruzdev]

Remove task_execution_id column from task_instances table

Summary

The task_execution_id column in task_instances is redundant. Since the composite primary key (instance_id, task_position) is the true identity for a task execution, task_execution_id should be removed from the database schema and all layers of the stack.

Problem

In MODE 3 (Kafka), task_execution_id is already derived from instanceId + taskPosition — confirming that the composite key carries all the identity information and the column is purely redundant.

Keeping task_execution_id around:

• Wastes storage and index space (idx_task_instances_task_execution_id)
• Misleads API consumers into thinking it's a stable identifier
• Adds unnecessary complexity to triggers, mappers, and ingestion code

Scope of changes

Database (MODE 1 — PostgreSQL)

• New Flyway migration: drop task_execution_id column and idx_task_instances_task_execution_id index
• Update normalize_task_event() trigger function (both fast path and FK-violation fallback) to remove task_execution_id from INSERT and ON CONFLICT clauses

Domain model

• TaskExecution.java — derive id from instanceId + taskPosition at the domain/mapper level (no database column needed)

JPA / PostgreSQL storage (MODE 1)

• TaskInstanceEntity.java — remove taskExecutionId field, update equals(), hashCode(), getId()
• TaskInstanceEntityMapper.java — remove taskExecutionId ↔ id mappings
• TaskExecutionJPAStorage.java — remove getTaskExecutionId reference

Kafka ingestion (MODE 3)

• TaskExecutionProcessor.java — remove generateTaskExecutionId()
• Mapper.java — remove generateTaskExecutionId() and its call
• task-instance-upsert.sql — remove task_execution_id from INSERT columns

Elasticsearch (MODE 2)

• task-events.json index template — remove taskExecutionId field mapping

GraphQL API

• WorkflowInstanceGraphQLApi.java — update getTaskExecution(id) identity strategy

Tests

• WorkflowInstanceGraphQLApiTest.java — remove setTaskExecutionId() calls, update assertions

Documentation

• Update CLAUDE.md (remove "Don't remove task_execution_id" guidance, update field mapping table)
• Update relevant docs in data-index-docs/, data-index-storage-postgresql/README.md, etc.

Identity strategy (resolved)

TaskExecution.id will be derived from instanceId + taskPosition at the domain/mapper level — the same derivation MODE 3 (Kafka) already uses today. No database column needed; the composite primary key (instance_id, task_position) is the source of truth, and id is computed on read.