In traditional WT file layout for a page on disk, the btree page header came first, then the block header, and they were both predictable sizes.  In disaggregated storage, for deltas, the delta header replaces the block header (but of course it may be different size).  Obviously this makes for some weirdness to find the block header (it's possible now because we always know what kind of thing (delta/full-page) we're looking at, but we want a stored item to be easily self identifiable).

We should shuffle the headers around to avoid this problem. (It also results in us having to maintain two sets of encrypt/compress flags.)