Improve HST reporting / diagnostics about the processing of a request

I was hoping we could store request statistics on some thread
local object (and every search that is finished add some object like
how much time the search took), and then in the clean up valve

1) clean the object again
2) log debug : how long every separate search took
2) log info about for example how many searches there where for the
entire page (and how long all of them together took)
3) log warning when number of searches for 1 pages exceeds some
configurable limit, or, when total search time for 1 page exceeds some
configurable limit

Perhaps it should not be done with thread local at all, but seems to
me simple solution. However, this is typically something that Woonsan
is most likely best of doing and building