feat: Unify Feature Transformations and Feature Views

[franciscojavierarceo]

What this PR does / why we need it:

This PR introduces a unified transformation system that consolidates the transformation capabilities into a single, powerful @transformation decorator. This addresses fragmentation across different transformation approaches and enables seamless training-serving consistency through dual registration.

from feast.transformation import transformation

@transformation(
    mode="python",           # Execution mode: python, pandas, sql, etc.
    when="on_write",         # Timing: on_read, on_write, batch, streaming
    online=True,             # Enable dual registration for training-serving consistency
    sources=[driver_source], # Input data sources
    schema=[                 # Output schema definition
        Field(name="trip_duration_minutes", dtype=Float64),
        Field(name="trip_cost_usd", dtype=Float64)
    ],
    entities=[driver],      # Entity relationships
    name="trip_features"    # Optional custom name
)
def trip_feature_transform(inputs):
    """Calculate trip features from raw driver data."""
    return [
        {
            "trip_duration_minutes": inp["trip_seconds"] / 60,
            "trip_cost_usd": inp["distance_km"] * 2.50 + 3.00
        }
        for inp in inputs
    ]

Which issue(s) this PR fixes:

#4584, #5716, #5689

Misc