Prevent methane lifetime reaching zero

[chrisroadmap]

in some very warm runs in e.g. ssp585 beyond 2100, using ch4_method="Thornhill", the methane lifetime adjustment due to temperature can get large. This causes this line:

np.log(1 + temperature * ch4_lifetime_temperature_sensitivity)

to become negative in the log term if ch4_lifetime_temperature_sensitivity is negative and temperature is large. This tends to happen in ensemble members above 15 K. The likelihood of FaIR being a reliable model in such extreme warming is small and these sorts of projections should be used with caution anyway, but there should be a sensible minimum value for the methane lifetime that can be set in the call to FaIR that would override the calculation if this log term becomes too small or NaN.