Do not split stored exit policy summary into type(accept/reject) and portlist.  At least not just yet

svn:r16553

svn:r16552

svn:r16551

svn:r16550

svn:r16549

svn:r16548

svn:r16545

svn:r16544

Having a reject-all policy summarized with a lack of summary is stupid.  Explicitly mention them as reject 1-65535.

svn:r16543

testsuite: Now that we add (pretty useless, but still) routerinfos to the routerlist for every one of our routerstatuses in the votes we pass again.  yay

svn:r16542

svn:r16541

svn:r16540

svn:r16539

svn:r16538

svn:r16537

svn:r16536

svn:r16535

svn:r16530

svn:r16529

svn:r16528

svn:r16527

svn:r16526

svn:r16525

svn:r16524

svn:r16523

svn:r16522

svn:r16521

svn:r16520

svn:r16519

svn:r16518

svn:r16516

svn:r16515

svn:r16514

svn:r16513

svn:r16512

Updated to remove dropping of failing guards and just focus
on the specifics of recording, storing, and learning
circuitbuildtimeout parameters.



svn:r16511

svn:r16510

automaphostsonresolve config option, because nroff/troff are
just too much like text.


svn:r16507

Describe how clients should use the exit summaries, what they may use a locally cached descriptor for, and that enclave exiting is still allowed

svn:r16501

svn:r16500