FIR with duration

[behinger]

Allows to specify duration for each FIR event.

Warning: predict + LMM is not really supported right now and we don't throw a warning. I documented it in FIRbasis and firkernel, but it is not easy to throw a warning.

Not sure how to proceed. It is not easy to fix, because the basis function will have variable duration, depending on the duration provided in the kernel. We could "cut" the kernel...

Why does HRF not have this issue?

Good question! Because we specify the max length for the HRF :)

This is how it looks like

[behinger]

please check, unclear if compatible with the new fastpath
time_expand_allBasesSameCols
in designmatrix.jl

[github-actions]

[JuliaFormatter] _{reported by reviewdog 🐶}

Suggested change

	# defining a firkernel with a duration = 1 sample, should be equal to the non-duration one
	f_dur = Unfold.firkernel([103.3;1],range(-0.1,step=0.01,stop=0.31))
	f_fir = Unfold.firkernel(103.3,range(-0.1,step=0.01,stop=0.31))
	@test f_dur == f_fir
	# test duration for samples = 4
	f_dur = Unfold.kernel(firbase)([1,4])
	@test all(sum(f_dur,dims=1) .== 4)
	# defining a firkernel with a duration = 1 sample, should be equal to the non-duration one
	f_dur = Unfold.firkernel([103.3; 1], range(-0.1, step = 0.01, stop = 0.31))
	f_fir = Unfold.firkernel(103.3, range(-0.1, step = 0.01, stop = 0.31))
	@test f_dur == f_fir
	# test duration for samples = 4
	f_dur = Unfold.kernel(firbase)([1, 4])
	@test all(sum(f_dur, dims = 1) .== 4)