#39 nav/march round mobility inconsistently

[Markus Armbruster]

Nav or march work with floating-point mobility, and
round when the movement ends. They round towards zero.
This rounds mobility cost up when the final mobility
is positive, and down if it's negative.

Exploit: you can save up to one point of mobility per
sector travelled.

[Markus Armbruster]

Logged In: YES
user_id=596270

`info Mobility' says:

If the mobility cost for something is a fraction, for
example 2.4,
then there will be a percentage chance that the actual
cost is rounded
up or down. In our example, there would be a 60% chance
that 2.4 is
rounded down to 2 and a 40% chance that it is rounded up to 3.

This makes sense, but it doesn't match the code.

I suggest to create a function

int <foo>_charge_mobil(double mcost)

for each kind of <foo> that has mobility, with behavior
matching `info Mobility', and use those functions
exclusively to charge mobility.