Make read_records return deterministic hash

[ilongin]

read_records() creates a temp dataset with a random UUID name. The starting checkpoint hash is derived from this name, making it different on every run — so checkpoints can never be reused for read_values, read_pandas, read_hf, and direct read_records calls with list data.

This adds deterministic content-based hashing: when read_records receives a concrete list (not a generator/iterator), a SHA256 hash is computed from the actual data and set as the starting checkpoint hash, overriding the random temp dataset name. read_values similarly computes a hash from its keyword arguments. This makes checkpoint reuse work for read_values, read_pandas, read_hf, and read_records with list input.

Generator/iterator-based inputs (e.g. read_database, read_records with a generator) still produce non-deterministic hashes since the data is consumed lazily and can't be hashed without buffering it all in memory. - in order to fix this we would need change how we approach to hash calculation in DataChain as currently it can be calculated without applying steps, but in order to support generators we need to move hash calculation into applying steps itself.

[cloudflare-workers-and-pages]

Deploying datachain with    Cloudflare Pages

Latest commit:	98fcc8e
Status:	✅  Deploy successful!
Preview URL:	https://ce1bdbc6.datachain-2g6.pages.dev
Branch Preview URL:	https://ilongin-1636-deterministic-h.datachain-2g6.pages.dev

View logs

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!