feat: add consistency metric - cpfd

[SeegerM]

Adds a new consistency metric based on weighted partial functional dependencies (pFD). Similar to related work, such as Pipino, but divided by the sum of the weights of the pFDs.

• Adds consistency_cpfd, a fully-automated consistency metric based on "Automatic Consistency Assessment through Partial Functional Dependency Mining" (Seeger, Papenbrock, Naumann, Ehrlinger — VLDB QDB'26). The metric aggregates a set of partial FDs into a single score in [0, 1] as $\mathrm{cpfd}(\mathcal{F}) ;=; \frac{\sum_{f \in \mathcal{F}} \omega(f)\cdot r(f)}{\sum_{f \in \mathcal{F}} \omega(f)}$, where ω is the gpdep-based genuineness weight and r = 1 iff the partial threshold ρ = 1 otherwise r = 0 .
• Adds a pfd data-profile task: new PFDImporter parsing the partial-HyFD output format [table.csv.col1, …]->table.csv.col_n#<partialScore>#<weightScore> (inline JSON also supported), and new DataProfileManager.store_pfd / get_pfds helpers
• Wires the metric into the registry and documents the new import format.
• pFDs are expected to be computed externally with the Java HyFD extension from SeegerM/cpfd-reproducibility and imported via the standard data-profile loader