Improve rbind() behavior on iterators

[oleksiyskononenko]

Currently we have quite strange rbind() behavior when the argument type is an iterator. For example, when we have a test.csv file with the following content

c1, c2, c3
1, 2, 3

read it with iread() multiple times and want to rbind the results, a typical use-case for iread(), this is what happens

>>> from datatable import dt
>>> it = dt.iread(["test.csv", "test.csv"])
>>> DT = dt.rbind(it)
ValueError: Cannot rbind frame with 3 columns to a frame with 0 columns without parameter force=True

The error message is not very helpful. In some cases (cannot reproduce it now) the message even looks like this

ValueError: Cannot rbind frame with 1152921504606846976 columns to a frame with 0 columns without parameter force=True

We should either support frame iterators directly in rbind() (the preferred way), or explicitly say they're not supported.