This reverts commit a5027dc7.

* t2:
  X Unexport SetXXX from packages API

* t2:
  X No longer rely on ZODB cache invariant for invalidations
  X ΔFtail: Keep reference to ZBigFile via Oid, not via *ZBigFile

Move all aspects of what is tracked into ΔFtail.
It is more robust if ΔFtail does not rely on properties of outside
zconn.Cache in order to work correctly.
It also potentially allows to move ΔFtail out of zdata package.

And it is even more memeory efficient, as before it was

	Connection.cache[zoid] -> ZBlk(ghost)->inΔFtail

and now it is

	ΔFtail.trackSetZBlk[zoid] -> zblkTrack

becuase Persistent embedded in ZBlk even in ghost state occupies alone
occupies ~ 16 words.

Now we keep only what is needed to be kept for tracking.

*ZBigFile is live object associated with particular ZODB Connection.

- gen-δtail is not used - we are building ΔFtail from scratch.
- gogcccrash did its work and the problem is now understood.

* t:

Now both normal and random tests for all Update and rebuild could be
exercised as part of regular test runs.

* t2:
  X rebuild: tests: Random testing
  X rebuild: tests: Don't exercise keys from keys2 that already became tracked after Track(keys1) + Update
  X rebuild: tests: Inline _assertTrack
  X rebuild: tests: Don't compute keyCover in trackSet
  X rebuild: tests: Don't recompute trackSet(keys1R2) several times
  X rebuild: tests: Move ΔBtail.Clone test out of hot inner loop into separate test
  X tests: Factor-out tree-test-env into tTreeEnv
  X xbtree: Less copy/garbage in RangedKeySet ops
  X rebuild: tests: Precompute kadj10·kadj21
  X rebuild: tests: Don't access ZODB in xtrackKeys
  X rebuild: tests: Don't access ZODB in XGetδKV
  X rebuild: tests: Don't reflect.DeepEqual in inner loop
  .
  .
  .
  .
  .
  .

Rework py allstructs to emit just set of random trees for one kv.
Rework TestΔBTailAllStructs to iterate through such obtained sets and
generate all pairs and triplets. Previously it was only pairs which is
not enough to verify rebuild properly.

X rebuild: tests: Don't exercise keys from keys2 that already became tracked after Track(keys1) + Update

58s -> 14s  (in -short mode)

t.Helper() is relatively expensive

61s -> 58s

63s -> 61s

68s -> 63s

73s -> 68s

We will need this not only in test update/rebuild.

rebuild tests ~75 -> ~73s.

82s -> 75s.

Use preloaded data, similarly to previous commit.

95s -> 82s.

Preload data when tTreeCommit is created and use that data in
the inner loops instead of accessing ZODB from those inner loops.

130s -> 95s

154s -> 130s

* t2: (50 commits)
  .
  X rebuild: Serial tests now started to PASS draftly
  .
  X rebuild: Don't return nil for empty ΔPPTreeSubSet - that leads to SIGSEGV
  X treediff: Fix BUG while computing AB coverage
  X ΔBtail.Clone had bug that it was aliasing klon and orig data
  X Fix rebuild with ø @at2
  X Fix bug in PPTreeSubSet.Difference  - it was always leaving root node alive
  .
  .
  X found why TestΔBTailAllStructs was not effective to find δtkeycov bugs
  .
  .
  X wcfs: assert that keycov only grow
  .
  X another bug in δtkeyconv computation
  .
  .
  .
  .
  ...