Use DiskMRU in loading MDEventWorkspace

[Janik Zikovsky]

... in order to keep a cache of boxes in memory to speed up processing. Refs #3445.

[Janik Zikovsky]

In [13638]:

Refs #3467: Steps to use DiskMRU, checking in to work on something else.

[Janik Zikovsky]

In [13639]:

Refs #3467: Fix windows test build

[Janik Zikovsky]

In [13660]:

Refs #3467: Try to fix the Windows build by purging any reference to TestHelpers and DataObjects from API.

[Janik Zikovsky]

In [13661]:

Refs #3467: Fix build in API.

[Janik Zikovsky]

In [13687]:

Fixes #3445: DiskMRU class should now have all the necessary functionality. Refs #3467: Implemented diskMRU into MDEventWorkspace, which now caches a certain amount of events in memory when using the file back end. First level of writable cache (can't change the # of events in a list yet).

[Janik Zikovsky]

In [13688]:

Refs #3467: Fix failing test temporarily since I have no idea what changed to make this fail.

[Janik Zikovsky]

In [13788]:

Refs #3504, #3467: Moved the DiskMRU class and related objects from API to Kernel.

[Janik Zikovsky]

In [13860]:

Refs #3504, #3467: MDBox will relocate blocks when saving.

[Janik Zikovsky]

In [13868]:

Refs #3467: The DiskMRU cache properly extends the size of the event_data field in the MDEW NXS file.

[Janik Zikovsky]

In [13897]:

Refs #3532: SaveMDEW has option to update the file-based back-end. When calling addEvent() on a disk-cached MDBox it does not require loading data from disk. Support for splitting a MDBox into a MDGridBox when using the DiskMRU. Refs #3467.

[Mathieu Doucet]

21 new warnings with the last commit?!

[Owen Arnold]

LoadMD works with DiskMRU from box controller.

[Stuart Campbell]

This ticket has been transferred to github issue ​4314