Add LLM banking assistant sample

[mxsisa]

Sample: Dozer + LLM + Vector database + LangChain banking assistant
This PR adds a runnable sample to getdozer/dozer-samples for the hyper-personalized chatbot use case described in the Dozer blog post about LLMs, vector databases, and LangChain.

What’s included

• A new sample at usecases/llm-banking-assistant

• dozer-config.yaml sourcing multiple banking datasets from local CSV files

• A LangChain app that:
  fetches customer context from Dozer REST APIs
  retrieves credit card knowledge from a Chroma vector database
  sends both into an LLM for personalized recommendations

• A reviewer-friendly --dry-run mode that prints the assembled context without requiring an API key

Datasets included:

• customers.csv
• customer_products.csv
• spending.csv
• transactions.csv
• credit_cards.csv

Dozer views
The sample uses Dozer SQL to build:

• eligible_cards
  joins customers to cards using annual_income >= min_income

• spending_summary
  aggregates spend per category

• transaction_summary
  aggregates transactions per merchant category

Why this addresses the issue:
Issue #1690 asks for a complete working sample based on the blog post’s banking chatbot idea, with:

• multiple datasets such as customer profiles and transactions
• Dozer APIs integrated with LangChain
• a vector database used for product knowledge retrieval
• a credit-card recommendation scenario
  This sample implements exactly that shape while keeping the setup lightweight for local review.

Implementation notes:

• Chroma is used as the vector database to keep the example easy to run locally.
• The Python app supports --dry-run so reviewers can inspect the assembled Dozer context without needing an OpenAI API key.
• The app lazy-loads optional dependencies so missing packages fail with clear setup guidance instead of import-time crashes.

Files added:

• usecases/llm-banking-assistant/README.md
• usecases/llm-banking-assistant/dozer-config.yaml
• usecases/llm-banking-assistant/requirements.txt
• usecases/llm-banking-assistant/app/chatbot.py
• usecases/llm-banking-assistant/data/*

Docs update:

• Added the new sample to the root dozer-samples README.

Limitations
I could not run a full end-to-end smoke test in this environment because dozer is not installed locally here.