#132 optimization for cache lookup protocol

[Scott Venier]

I was digging through '-cdebug All' output and I
noticed man calls that followed this pattern:

13:01:14.818 11/01/2006 -- CALLING -- FreeVariables
(call #36)
pk = 84bf0fdb125d1c86 618377fb60aecea2

13:01:14.819 11/01/2006 -- RETURNED -- FreeVariables
(call #36)
epoch = 0
isEmpty = false
names =
names.tbl = <empty>

13:01:14.819 11/01/2006 -- CALLING -- Lookup (call #37)
pk = 84bf0fdb125d1c86 618377fb60aecea2 epoch = 0
fps =
<< no fingerprints >>

13:01:14.819 11/01/2006 -- RETURNED -- Lookup (call #37)
result = Hit
index = 2
value =
fp = eee4d176f590d664 c74a852e0a4c0c3a
DIs = {
0x45f872f1, 0x6d02802d, 0x4fd1cf82, 0x59929c0e,
0x4771cf82 }
tbl = <empty>
bytes = 00000002 00030003 0004666f 6f310003
00020004 ... (823 total)

Specifically, the FreeVariables call returned isEmpty =
false and names.tbl = <empty>. This means that the
Lookup call is definitely going to result in a hit.

In this case the FV call could return the hit
information directly, saving an extra network round
trip to the cache.

In this particular build (against an empty cache), 130
out of 373 FreeVariables calls followed this pattern.