ConsIndShockSolver is always chosen now, and calc_bounding_values() works with new distribution format.

Perfect foresight model now produces a conditions_report field in addition to logging. Might need to check on the spacing with logging.

Not done yet. Also fixed a couple small mistakes in PF.

Still need to write "verbose" comments for check_conditions.

Still lacks Harmenberg growth patience factor/condition, as well as comments about the Modigliani mortality adjusted GIC. Will look at paper more closely.

There were a few typos. Violating FVAC but not WRIC now produces an ambiguous message, as the solution only *might* not exist.

Old add_stable_points methods in the PF and IndShock solvers were being run *every* period, which is very costly and essentially pointless. The new method lives on the IndShockConsumerType class and checks for relevant conditions before evaluating.

I need to double check, but I think the solution was accelerated significantly by getting rid of that code.

Search for mNrmTrg even if GICMod fails. Also put mNrmBal and mNrmTrg into top level. Need to change later. These changes were made while updating the BST dashboard notebook to be compatible.

The conditions_report is now sent to _log.info if not quiet. ConsIndShock.post_solve will now run calc_stable_points() if appropriate, so that the target and balanced-growth mNrm levels are added to the solution.

At the default parameters (infinite horizon), the main branch takes 0.328 seconds to solve on my computer. After moving the calc_stable_points code outside of the solver loop and only running it at the end (if appropriate), it takes 0.100 seconds to solve. Over two-thirds of solution time was being *wasted* on calc_stable_points, whose work is not used during the solution and *could not* plausibly be used.

Per SB's request, I'm splitting the change to calc_stable_points to be in a separate PR. This commit reverts the changes from a prior commit and should make the tests run properly now (because mNrmStE has not been renamed).

A new dictionary has been added to PerfForesightConsumerType and IndShockConsumerType, called auxiliary. It contains all the various patience factors and other semi-useful values that are constructed for check_conditions, but aren't needed to solve (nor simulate) the model. This dictionary can be renamed with a simple replace-all on self.auxiliary.

Patient factors and conditions report now life in the bilt dictionary.

With log utility, the expression for the utility-compensated permanent shock (and thus the value of autarky factor) would break with a divide by zero error. This is now fixed, with the end result that the VAFac is just beta with log utility.