bugs in estimation of length

[nukappa]

bugs found by simulations:

• when the tail range does not include the simulated value then all probabilities computed are NA
• when the precision in tail range is such that the simulated value is not in there, the estimated length is rounded up to the next one, even if it's much further away than the previous value (ceiling vs round somewhere?)

[mschilli87]

I think (at least) the 1st one was solved by now.

[nukappa]

i think not yet.
for a tail range of (10000, 15000, 1000) it gives finite probabilities, which is wrong, see #83.
for a tail range of (1, 5, 1), with a read non reachable by any insert size, it breaks down due to division by zero