But I will think about decreasing gate resistors a little bit. Now 10R is used so small bannana shape is visible on gate charging and discharging. Or maybe I will add diode+resistor in parallel just to turn it off faster and keep slower turn on time because of EMI...
I've seen higher gate resistances working just fine - depends on the MOSFETs and it's just the usual power dissipation vs. EMI tradeoff game 
Yes, I know. I have seen even 100R in ST BLDC controller

Worked well with 6-step but it was a tragedy when I was trying to implement sinusoidal driving and had to be reduced

So it also depends on control scheme and your needs. But I am also aware that high rise times look good on DSO but are your enemy in EMI

But on other hand additional inductance should add some slope

I have just checked that driver already has 5R of internal resistance for LOW TO HIGH transition and 0.55R for HIGH TO LOW transition so it is already asymmetric. I will leave that for further testing in the future

And what is most important current ripple significantly decreased with 50kHz PWM ;-) I wonder what will be the effect of additional chokes in serie, it should be even better ;-)
Business as usual 
I have added 2x 4,7uH in series with motor windings and just as expected, current ripple has been decreased. It is not enormous change, something about 25%, but always something is better than nothing

Does it mean that effective inductance of motor is about 37,6uH? I think so and it sounds quite realistic. It would be 36,5% of 103uH presented in datasheet and they were talking about 30-80% for calculations..