fsck takes forever with a giant file

[EtchedPixels]

If we get a giant file we start walking all the blocks one by one patiently reading direct and indirect blocks to ascertain that yet again the block is a gap.

What we should do is spot that either the first or second indirect is 0 and skip on 256 blocks or 64K blocks by making getblkno also return the next viable candidate block to inspect.

[EtchedPixels]

Partly fixed by spotting re-reads of the same block. Really we want a 2 block cache