feat: Execute on Ibis backend via Narwhals

[zilto]

Prework

• Read and abide by the Pointblank code of conduct and contributing guidelines.
• Search for duplicates among the existing issues (both open and closed).

Proposal

Currently

Currently, Pointblank handles Ibis tables and columns by converting them to polars or pandas. This means:

1. Data is on backend (DuckDB, Snowflake, etc.)
2. Data is loaded into memory via Ibis to Polars
3. Checks are evaluated eagerly in-memory via Polars
4. Checks report is available in-memory

Proposed feature

#192 requests LazyFrame support for Polars lazy execution. Once that's supported, it should be trivial to add lazy support for Narwhals. Via the recent Narwhals-Ibis integration, it should be possible to execute checks on-backend. Meaning we would have:

1. Construct a query plan from all lazy checks via Narwhals (polars-compatible API)
2. Pass the Ibis Table reference to the query plan
3. The query plan is executed on-backend via Ibis; the full dataset never leaves memory
4. The query plan results are loaded into memory (i.e., checks report)

This Narwhals-Ibis bridge enables on-backend compute for most features (e.g., DataScan)