Insert added-backed samples at the right sampling times

[szhan]

In #193, we are adding back excluded samples from the past N consecutive days to the trees. But even though these samples are not collected on the same date as the newly added, non-excluded samples, their nodes have the same times as the newly added samples. To correct this, I think attach_tree needs to be modified.

[szhan]

Added-back sample nodes (all of which are terminal nodes in this case) are assigned to times equal to the time difference between the sample collection date and the current date of daily extension (in units of days ago).

For the internal nodes, their times are the max sample node time plus an arbitrarily chosen small value (e.g. 1e-6) multiplied by a positive integer, which is incremented every time an internal node is visited during postorder traversal.

I think this should work for now. The small value of 1-e6 should work unless we have over a million internal nodes.

[szhan]

Having gone through the functions used to build daily trees or trees with added-back samples, probably we would want to redo it, especially once UPGMA is replaced with NJ tree or something, so that it is easier to adjust the times of the internal nodes. For now, it is probably okay.