Not all boards have an SD card.

This is a *breaking* change and requires a bootloader change.

Setting PWM_DISARMED to 0 results in no PWM output to the ESC for the pusher motor.
Most ESCs start beeping endless in short intervals if they don't get a signal.
I remove changing this parameter which is 900 by default to always command the motor to stand still.

Now that the lockstep changes are merged to master.

Now that this is merged to master in jMAVSim.

This is just because the branch got merged.

It seems that the hysteresis test fails every now and then, presumably
due to timing issues. The tests needs some improvements, e.g. isolating
it from the system time.

It turns out the noise of the airspeed depends on the least significant
bit of time which is now always 0 because time is quite regular with
lockstep. Therefore, we need a better source for the randomness. I chose
to use the sign of the y-gyro and it seems to work.

This should fix a race because the use of future/promise wasn't done in
a thread-safe way.

An include was missing.

This fixes a HITL bug and uses promise/future instead of sleeping.

This brings build fixes for Gazebo 7.

For the non-lockstep case we want to use CLOCK_MONOTONIC if possible.

This means that we can use the normal host scheduling (not lockstep) for
the unit tests.

In the future we might add unit tests using lockstep as well.

Instead of using the define __PX4_POSIX_SITL it makes more sense to have
a define just to determine if the lockstep scheduler should be used.

The -t argument had been deleted because I didn't realize that it was
actually being used. Therefore, this brings it back and we now use -c
for TCP.

This fixes all world files at 250 Hz simulation rate.

This brings support for TCP and lockstep.

This is not suposed to ever trigger anyway.

@bkueng reviewed this and found that it is dangerous to continue on
timeout because we don't update other gyro sensors in the timeout case.

According to @bkueng this was to schedule other processes but can now be
removed.

@bkueng found that the old implementation was likely to wrap-around
given seconds is only a uint32_t. We now cast it directly to uint64_t
and therefore should fix this problem.

The new board config changed everything, so I needed to move the
lockstep_scheduler repo.

Consistency over taste :(.