Better documentation of `remake` mutating original problems

Normally remake does not affect the input problems. However, for JumpProblem this is the case:

using Catalyst, JumpProcesses
# Creates the model.
rn = @reaction_network begin
    p1*p2, A + B --> C
end
@unpack p1, p2 = rn

u0 = [:A => 1, :B => 2, :C => 3]
ps = [:p1 => 3.0, :p2 => 2.0]
dprob = DiscreteProblem(rn, u0, (0.0, 1.0), ps)
jprob = JumpProblem(rn, dprob, Direct())

jprob.massaction_jump.scaled_rates[1] # Returns 6.0.
jprob2 = remake(jprob; p = [p1 => 0.0]) 
jprob.massaction_jump.scaled_rates[1]  # returns 0.0 
jprob2.massaction_jump.scaled_rates[1] # returns 0.0

This should be documented properly, and a kwarg added to instead make a deepcopy of the input problem (so that there is no mutation).

Start with the remake entry point and the JumpProblem handling, then run the Julia reproduction in the issue to observe mutation of massaction_jump.scaled_rates. Done means documenting the current behavior and adding a keyword option that prevents mutation by deep-copying the input problem.