Reduce max_rows on PI requests

[einarsi]

Issue added on GitLab by @knudsvik 2018-06-12:

Based on experience, 100.000 rows got a time-out.

[einarsi]

Proposed solution: Wrap parts of _read_single_tag() in a try-except-clause where we try to fetch data n times until we report a failure and die. max_rows gets automatically reduced (halved?) on each attempt with feedback to user so he/she can adjust manually next time. Also let max_rows be parameter to IMSClient creation.

Until then, set max_rows manually:

c = IMSClient('myserver', 'pi')
c.handler.max_rows = 50000