Add 5 database connectors, fix SQL injection, and standardize agent result models

[yelinso344]

Summary

• Add MongoDB, Amazon Redshift, DuckDB, Microsoft Fabric, and Elasticsearch connectors
• Fix SQL injection vulnerability across all SQL connectors
• Standardise PDF agent result classes to Pydantic BaseModel (consistent with CSV agents)
• Fix Elasticsearch body= API breakage (removed in v8+)
• Install missing databricks-sql-connector package
• Update requirements.txt, README.md, and .env.example to reflect all changes

Changes

New connectors (src/connectors/)

• mongodb.py — URI-based, returns collection as DataFrame
• redshift.py — uses redshift-connector, mirrors PostgreSQL interface
• duckdb_conn.py — local file-based, uses .df() native return
• fabric.py — ODBC Driver 18 for SQL Server (OS-level dep, noted in README)
• elasticsearch_conn.py — optional auth, SSL toggle, fetches index as DataFrame

Security fix — SQL injection (src/connectors/)

Added _validate_identifier() to postgres.py, mysql.py, databricks.py,
snowflake_conn.py, and bigquery.py. Rejects any identifier containing
characters outside A-Za-z0-9_. before SQL is executed.

Pydantic standardisation (src/agents/)

Converted PDFParserResult, EntityExtractorResult, RiskDetectorResult,
and ActionExtractorResult from hand-rolled classes with manual model_dump()
to pydantic.BaseModel — consistent with all CSV pipeline agents.

Bug fixes

• elasticsearch_conn.py — replaced deprecated body= kwarg with direct params (v8+ API)
• databricks.py — databricks-sql-connector was missing from venv and requirements

Docs & config

• README.md — updated connector status table, architecture tree, code examples,
  roadmap, tech stack, contributing guide, and agent pattern snippet
• .env.example — added entries for all 10 connectors
• requirements.txt — regenerated from venv (adds pymongo, redshift-connector,
  duckdb, elasticsearch, pyodbc, databricks-sql-connector)

Test plan

• Run pytest tests/ -v — all 11 existing tests pass
• CSV pipeline — upload demo CSV, run all 5 agents
• PDF pipeline — upload demo PDF, run all 5 agents
• MongoDB — connect to local instance, fetch a collection, run pipeline
• Redshift — connect, fetch table, run pipeline
• DuckDB — point at a .duckdb file, fetch table, run pipeline
• Elasticsearch — connect to local instance, fetch an index
• Microsoft Fabric — requires ODBC Driver 18 installed at OS level
• Verify SQL injection guard rejects a table name like users; DROP TABLE users

[harshitboots]

This PR was created from the same base commit as PR #2 (DuckDB), and likely PR #3 (Redshift). Since then:

1. Rebase onto main
2. Resolve Conflicts
3. Force Push the Rebased Branch

kindly rebase it seems ok to merge