Add additional example tasks (Titanic, Iris, Wine Quality, House Prices, Credit Default) to validate pipeline

Summary

Add several small, public datasets as tasks to ensure the multi-agent pipeline works across regression and classification scenarios (binary + multiclass) and different feature types.

Proposed Tasks

1. california-housing-prices (already present) – baseline regression
2. house-prices-advanced (Kaggle House Prices) – richer regression with categorical encoding needs
3. titanic-survival – binary classification
4. iris-classification – small multiclass classification
5. wine-quality (red) – regression with mixed numeric distributions
6. credit-default (UCI / Kaggle variant) – binary classification with class imbalance

(Optionally later: adult-income, mnist-tabular (flattened), forest-covertype.)

Directory Layout (example)

machine_learning_engineering/tasks/ titanic/ train.csv test.csv task_description.txt iris/ train.csv test.csv task_description.txt ...

Example task_description.txt (regression)

task_name: house-prices-advanced
target: SalePrice
id_column: Id
metric: rmse

Example task_description.txt (binary classification)

task_name: titanic
target: Survived
id_column: PassengerId
metric: f1
problem_type: classification
num_classes: 2

Example task_description.txt (multiclass)

task_name: iris
target: species
id_column: id
metric: accuracy
problem_type: classification
num_classes: 3

(If id column not in original dataset, synthesize one.)

Acceptance Criteria

• Each task runs end-to-end with run_pipeline.py without code modifications.
• Generated workspace folders contain predictions without errors.
• Metrics calculation does not crash (classification vs regression handled).
• README gains a short “Available Example Tasks” section.
• Optional: lightweight smoke test added (pytest marker) for at least 2 tasks.

Implementation Notes

• Normalize column names (snake_case) if needed.
• Ensure train has target; test omits target.
• Keep CSVs small (<200KB each) to avoid repo bloat.
• Add a data README citing original sources + licenses.
• If different metrics required (e.g., log rmse), document future extension.

Checklist

• Datasets vetted for license
• task_description.txt for each
• Updated README
• Added smoke tests
• Verified pipeline logs clean