Proposal: add point-in-time momentum alpha model to v2

Proposal

Add a deterministic momentum AlphaModel to the v2 signal layer.

Initial scope

• Configurable trailing-return horizons (for example 21, 63, and 126 trading days)
• Optional short-term skip period to reduce reversal exposure
• A conviction value bounded to [-1, +1]
• Neutral output when price history is insufficient
• Transparent return components and parameters in the Signal
• Synthetic tests covering bullish/bearish/flat series, insufficient history, missing observations, conviction bounds, and explicit prevention of future-data leakage
• Package export and one example strategy configuration

I plan to keep the first PR limited to the model, tests, export, and example configuration, with no execution or UI changes. I did not find an overlapping open PR.

Is this direction aligned with the current v2 roadmap, and are there preferred momentum horizons or normalization conventions I should follow?