ENH: read_csv to support start/stop

[jreback]

would be useful (mainly for out-of-core) to be able to specify the start and stop rows and quickly skip to them to get a specific chunk out of a file.

these rows refer to data rows / alternately this could just return the nth chunk
(but skip parsing the n-1 chunks for efficiency)

[makmanalp]

read_csv has a skiprows and nrows which almost does this. The problem is that I think header is counted from after the skiprows row.

[WillAyd]

Believe this can be handled as noted above. If that's missing the point please reopen