Performance refactorings

[ericpohl]

Cache intermediate results to avoid excessive file I/O and XML parsing. This is particularly noticeable when doing a quick read to get the metadata from a LoggedData object (grower/farm/field, operation type, crop, date etc).

Refactorings break down as follows:

• Cache the directory listing in ExtensionMethods.GetDirectoryFiles. Currently GetDirectoryFiles enumerates all files on every call to find a single file and can be called many times for  ISOTimeLog.GetTimeElement (each looking up a TLGnnnnn.xml), plus additional calls for .bin files in  TimeLogMapper.ReadTimeLog and FillDPDGeometryDefinitions. Every call re-enumerates the same directory. Create a simple dictionary to hold previous results, and clear at the end of each plugin.Import() and plugin.Export() call.
• Enumerate external XML files once in ISO11783_TaskData.ReadXML. ProcessExternalNodes is called for 15 different XML element prefixes. Each call potentially checks the directory for matching files, but most datasets only have external files for TSK.
• Cache parsed ISOTime results from ISOTimeLog.GetTimeElement. The method is called multiple times from 4 different callers; each call re-reads and re-parses the same TLGnnnnn.xml file.
• Build a mapping from DDI to Representation in Representationmapper. Currently RepresentationMapper.Map() does a linear LINQ scan over ~600+ representations on every DDI lookup. With many timelogs and 10-20 data elements each, it can be hundreds of scans.
• Pre-index DeviceProperties by DDI in DeviceHierarchyElement to avoid scanning DeviceProperties multiple times for different DDI's.