Whether or not you need a resistor + diode rather than a simple diode snubber depends on your application and the relay characteristics. If you are switching a large inductive load at close to the relay's contact rating, its likely to be beneficial. You may also need a RC snubber across the relay contacts. If you are switching a resistive low voltage load at a fraction of the relay's rating, its probably unnecessary.
Consider the case of a diode + matched resistor snubber for your relay. Assuming the Rds_on of the MOSFET is negligible compared to the coil resistance, you get nearly the whole supply voltage (12V) across the relay coil when on. As it draws 45mA, the coil resistance must be about 267 ohms. Lets assume you use a 270R resistor in series with the diode. Neglecting stray capacitance, if the MOSFET shuts off quickly enough, that 45mA gets rapidly diverted through the diode and resistor, and will rapidly develop a voltage drop of 11.8V across them, (IR + diode Vf @ 45mA) driving the drain to 23.8V then decaying exponentially as the stored energy in the coil is used up.
23.8V is near enough to twice the supply voltage, but you need some margin above that. e.g. if the supply is 10% high, it will peak at around 26.2V which is getting pretty close to the rating of a 30V MOSFET. If you have to cope with a max 16V supply expect it to peak at 31.7V which is above the rating of a 30V MOSFET. Of course you can trade off slower contact opening against peak voltage by decreasing the resistor, but as its the total loop resistance + the coil inductance that determines the LR time constant, even if you do away with the resistor its only a factor of two worse, so the benefits of having a lower value resistor are marginal.
The relay voltage spike is negative, so gets clamped by the diode to the diodes forward voltage.
Please explain your reasoning. Negative with respect to what? As 'Magic' explained, during the back-EMF spike, the MOSFET end of the coil will swing above the supply rail, not below ground.