Hey Jensa! Thanks for posting this question, I studied power electronics in grad school and this question touches upon some things I learned in my first power electronics lab class. First, I checked out the IRLL2705 datasheet available here: https://www.infineon.com/dgdl/irll2705pbf.pdf?fileId=5546d462533600a401535664653425db
This is a nice general purpose switching MOSFET. Good find, I may use it myself in the future! 
Regarding your question about the PWM switching application:
-
The total turn-on time (td-on + tr = 6.2ns + 12ns ~= 18ns) and turn-off time (td-off + tf = 35ns + 22ns ~= 57ns) is totally in line with industry standards for a silicon MOSFET of this switching class. It’s going to be hard to find something a lot better without moving to something exotic like GaN or SiC, and frankly why on earth bother? I’m not sure how you came up with “… something that can turn on/off faster than 20hz”. If you take the switching times from the datasheet and add them up (18ns + 57ns = 75ns), then compute the maximum theoretical frequency if the device were only switching on and off again as 1/(total switching time) you get 1/75ns = 13.33E6 or 13.33MHz. Realistically, in a generic PWM application you’re looking at switching at about 20kHz to maybe a few hundred kHz max. So assuming a switching frequency of 100kHz, the total PWM period is 1/100kHz = 10µs. So in this 10µs period, what percentage is spent switching on and off with the IRLL2705? 75ns / 10µs = 0.75%. So less than 1% of your cycle is spent switching. Perfect, there’s nothing wrong with this!
-
Your IRLL2705 will not switch this fast when connected to a 3.3V microcontroller, not anywhere near this fast actually. In fact, this may be why you’re seeing only 20Hz maximum switching if you tested this out. Why? Take a look at the the datasheet, figure 10a. This is the circuit used to test the switching times. Notice that the square wave driving the gate is an ideal voltage source of 10V amplitude, driving the gate through a very small (6.2Ω) resistor. So how much current is flowing out of this ideal source during a typical turn-on transition (td-on + tr = 6.2ns + 12ns ~= 18ns)? An approximation of the average current required over the entire 18ns can be easily computed. Look up Qg in the datasheet, it’s up to 48nC. Recall that 1C = 1A * 1s. So 48nC in 18ns = 48nC / 18ns = 2.667A of current! There’s no way that a microcontroller can drive this MOSFET directly if you’re wanting anywhere near the switching performance specified in the datasheet. While the MOSFET gate is separated by a very good insulating layer from the drain and source, the gate still has capacitance relative to these, and in power MOSFETs this capacitance is significant if you’re wanting to quickly charge and discharge it to turn the MOSFET on and off. I recommend driving the MOSFET with a gate driver IC like the the MCP1416 (even though the MCP1416 is only rated for 1.5A drive, it will still work fine for a lower switching frequency application).
Don’t forget to put a (~100kΩ) pull-down resistor on the input to the '1416 so that the MOSFET is off while your µC comes out of reset and your outputs are tri-stated!
One other tip: Make sure you have a schottky diode in your PWM circuit if your load has any kind of inductance or even any stray inductance to carry the current while the MOSFET is off. Have fun!