Design of persistent tag cache

[Greg Ward]

[me, on Matt's tag cache design]
>>   <headnode> <headrev> <tagnode>
>>   [...repeat: 1 line per head...]
>>
>>   <node> <tag>
>>   [...repeat: 1 line per tag...]
>>
>> Good things about this design:
>>   * it's easy to detect if the entire cache is still valid: if the set
>> of heads in the cache == the current repo heads, the cache is all good
>> and we don't have to read any .hgtags files

[Matt reminds me]
> We can do better. If the tip is stored correctly in the cache, we can
> assume all the other heads are correct.

Sure, but that's a second-order optimization.  Given the proposed
structure of the cache file, we still have to read the whole thing.
Comparing tips just means we can test "cacheheads[0] ==
currentheads[0]" and possibly skip comparing the whole lists.  And we
don't have to store all the cached heads in a list anywhere.  Compared
the cost of what we're *really* trying to avoid, this is small change.

> This ignores mq/strip, which
> will have to invalidate/update the cache for all the rules it breaks.

On reflection, I think that rollback is equivalent to strip from the
POV of tag caching. They can both destroy one or more changesets, they
both remove a head, and they can possibly expose older heads.

So I think that whatever I do to handle strip also needs to be done
for rollback.

Greg