Clients might need to to Interval math too, so move a simple Interval
class and start putting helpers in there.

This lets us use simple indexing to get timestamps from the table,
which allows us to use 'bisect' directly without needing a proxy class.

Implemented in command line, client, server, nilmdb, bulkdata

Previously, we ignored problems when sending intermediate blocks,
since getting more data might make the next attempt succeed.
But in practice, malformed data would just build up, causing
problems.  Raise an exception if there's too much data remaining
after trying to send an intermediate block.

strtod() and friends will happily swallow newlines, so we need to skip
over spaces and look for that sort of thing manually.

The bad line is printed out on a new line, and a third line
with a ^ to indicate the position of the error.

For rows and seconds only.  Extents still give None if they don't
exist.

Can show the set-difference between the interval ranges in two
streams.

Enable the following pragmas: synchronous=NORMAL, journal_mode=WAL.
This offers a significant speedup to INSERT times compared to
synchronous=FULL, and is roughly the same as synchronous=OFF
but should be a bit safer.

Also changes the HTTP parameter from "extent" to "extended",
and the commandline parameter from "extent" to "ext".

"import nilmdb" doesn't do much; "import nilmdb.client" or "import
nilmdb.server" is now required.