Consider a two stage solution using an open-drain inverter or comparator followed by a CMOS inverter.
The first stage could either be built with either a discrete transistor or an open-drain logic device. In either case, the rise time will be limited by the RC combination of the pull-up resistor and the load capacitance. With a fairly low-valued pull-up resistor, like say 2 kOhms, and a 15 pF load (for the next-stage chip I'll suggest below), you'd have a 30 ns time constant on the rising edge, which is reasonably within your specs.
For the second stage, the old-school CMOS CD4069 operates at up to 18 V, and has propagation delay of 30-60 ns and rise/fall times of 50-100 ns when operating at 10 V (improving for higher voltage).
It's not clear from your question if the 15 - 100 ns "fronts" you specify is for the rise time or the propagation delay. With this combination you meet the requirement for rise time; but may be near the limit for propagation delay -- you might have to adjust the pull-up resistor for the first stage to trade off power consumption vs delay.
After thinking about this some more, I realized this solution will probably not be fast enough to meet your specs, because the CD4069 rise-time spec is based on a 50 pF load. With 1000 pF load you are not likely to see 100 ns risetimes -- 1 us seems more likely.
An alternative may be to construct your own CMOS inverter from discrete FETs or a complementary pair like NTJD4158C. The benefit of using discretes is you will be able to simulate the performance of your switch and be sure it will give the rise/fall times and propagation delays you need before you build it.