Add a clearly licensed dataset replacement option

[sardorcodev]

Explore a replacement dataset with clear redistribution rights for future model training.

Scope:

• Find an agriculture/crop recommendation dataset with explicit license.
• Document source, license, schema, limitations, and checksum.
• Compare schema compatibility with the current training pipeline.
• Do not replace current model artifacts without release-gate approval.

Acceptance criteria:

• Candidate dataset is documented.
• License is clearly stated.
• Training compatibility is assessed.
• No unsupported redistribution claims are made.