Retroactively code review PR 602 (subproblem selection) if desired

I remain of two minds about having merged the branch without external review. The rationale for merging without review was:

Roll the feature out to students without further delay. The most useful feedback on it will really come from use, rather than inspection, and V. has been projecting the impression that giving the students a concrete implementation to port papers to is a high priority.
The subsystem is unstable, so there is actually an advantage to not having too many people invested in it.
The best reviewers would be Taylor and Anthony, and they are both quite busy.

The rationale for obtaining review was:

Code reviews are good.
The implementation is non-trivial and at least somewhat error-prone, so a second pair of eyes would have particular value.
The subsystem may need low-latency edits as students discover problems with it, so having one more person familiar with it confers an advantage.

For reference, here is how big a change it was,

$ git diff --stat origin/master...axch-poster-subproblem-selection
backend/lite/infer/__init__.py                    |  18 +-
backend/lite/infer/dispatch.py                    | 296 ++++++++++++++++++++++++++++++++
backend/lite/inference_sps.py                     | 550 ++++++++++++++++++++++--------------------------------------
backend/lite/orderedset.py                        |   4 +-
backend/lite/scaffold.py                          |   2 +-
backend/lite/trace.py                             | 232 +++++--------------------
backend/untraced/trace_search.py                  | 294 ++++++++++++++++++++++++++++++++
python/lib/engine/engine.py                       |   4 +-
python/lib/engine/inference.py                    | 181 +++++++++++++++++++-
python/lib/parser/venture_script/grammar.y        |  23 ++-
python/lib/parser/venture_script/parse.py         |  64 +++++++
python/lib/parser/venture_script/scan.py          |   2 +
script/vendoc                                     |   1 +
test/inference_language/test_advanced_blocking.py | 197 ++++++++++++++++++++++
tool/pylintrc                                     | 125 --------------
tool/pylintrc-1.1.0                               | 122 ++++++++++++++
tool/pylintrc-1.5.1                               | 125 ++++++++++++++
tool/venture-epylint.py                           |   5 +-
tool/venture_mode_font_lock.py                    |   1 +
19 files changed, 1553 insertions(+), 693 deletions(-)

though that does include some moving code around and fixing junk that was in my way without technically being related.

The code to review is PR #602, namely the axch-poster-subproblem-selection branch. There is a logical stopping point at the axch-subproblem-selection branch, which contains just the operations implementing selection, without syntactic sugar for them.

probcomp / Venturecxx

Retroactively code review PR 602 (subproblem selection) if desired #603