Calc: Further cache simulation results

[Noel Grandin]

On 2018/10/01 2:19 PM, Kohei Yoshida wrote:
>> On October 1, 2018 at 5:31 AM Michael Meeks <michael.meeks at collabora.com> wrote:
> 
>> 	Kohei poked at a chunk of profiles like this in the past; probably
>> worth getting his input; any thoughts ?
> 
> So, I believe when I reworked the internals of SvlListener & SvlBroadcaster to switch to unorderd_set back in 2013, I did consider using vector instead, but decided to stick with unordered_set.  I don't remember the exact reasoning behind 


For those who couldn't see the original discussion - the point of this was reducing the size of ScFormulaCell from 160 
bytes to 108 bytes, to win some performance in the calc engine.

I have dropped it for now, since the results were not universally great - make sc.perfcheck shows small gains across 
most of the results (1-5%), but a couple of large drops in performance (50% worse).

Another option would be to simply make the SvtListener an optional part of ScFormulaCell, since a lot of the time we are 
listening via the associated formula group.