state_start/state_end to define_kernel()

[levisc8]

Mentioned by @Aariq in levisc8/ipmr_esa#3 - I think this would make a more intuitive UI. The rationale is that state_start/state_end are part of the symbolic kernel defintion anyway (i.e. they're the "z',z" in P(z',z) = ...), so they should be attached to define_kernel(), rather than a separate function.

This probably means dropping define_impl() and moving int_rule to define_domains(), or vice versa (depending on which function name feels more intuitive).

define_kernel(..., state_start = "z", state_end = "x") %>% 
define_domains(x = c(L_x, U_x, n_x), z = c(L_z, U_z, n_z))

[Aariq]

I like define_impl(), for what it's worth. To me (someone relatively new to IPMs), it feels like defining the meshpoints is part of "implementation".