Add new flag to allow pre-allocating disk blocks

In some clouds the guest's system disk is backed up by disk image file(s) that are located somewhere in the host's file system. To save space, and/or to speed up how fast the VM is built (assuming images have to be copied and/or transferred), some clouds use thin-provisioned or thick lazy-zeroed images which in both cases delays block allocation of unused empty blocks.

This feature can have a performance hit (increase latency) when doing writes, since it will have to first allocate the blocks and might need to zero the new allocated blocks. In some clouds, for high block IO workloads it is recommended that users initially 'fill up' their disks by writing, and then deleting, a dummy file full of zeroes to make sure the blocks get allocated and later avoid the higher latency writes.

It would be helpful to include a flag to PKB that can expand a scratch disk to almost its full capacity, prior to executing an block IO benchmark.