Boundary issue for kernels

[alexpiet]

Right now, aligning events for the kernels are binned into fit_trace_bins, and then I construct the toeplitz matrix. However, this creates a boundary issue whereby events that happen outside the bins, but within a kernel length away from the bins are not influencing the design matrix, but they should be.

In order to fix this, I think I need to create a wider set of bins, and then trim the toeplitz matrix. This will only impact a small part of the design matrix, so its low priority.

[alexpiet]

The running and pupil kernels show the issue here