Add weighted samples for TRF

[Hugo-W]

Description

Enable weighted samples for TRF estimation, allowing for confidence-based weighting (e.g., from EOG or muscle artifact data) or phase information.

Tasks

• Implement support for weighted samples in TRF estimation.
• Allow weighting based on confidence of samples (e.g., from EOG or muscle artifact sliding window data).
• Support phase information as a weight matrix or to mask data points.
• Implement robust TRF using a log-based error function (e.g., $d(e) = \log(1 + (e/\sigma)^2)$) instead of classic MSE.
• Use iterative methods (e.g., conjugate gradient) for robust TRF estimation.

Mathematical Background

For weighted samples, the regression formulation is:
$$\min_{\vec{\beta}}||W(\vec{y} - X\vec{\beta})||^2$$
where $W$ is an $N \times N$ diagonal matrix with error weights on the diagonal. The solution is:
$$\vec{\beta}_{opt} = (X^TW^TWX)^{-1}X^TW^TW\vec{y}$$

Acceptance Criteria

• TRFEstimator supports weighted samples via a weights parameter.
• Robust TRF with log-based error function is available.
• Iterative solvers (e.g., conjugate gradient) are used for robust TRF estimation.
• Backward compatibility with existing TRFEstimator usage is maintained.

Priority

High