Improve test stability

[jpn--]

This pull request includes several changes to the activitysim/estimation module, primarily focusing on updating the larch version handling, modifying test cases, and adjusting model parameters.

The key change is applying a constraint on the trip model choice model, to solve an over specification problem and ensure the estimation results are stable.

The most important changes are grouped by their themes below.

Test Case Adjustments:

• Added a condition in the test_location_model function to lock a parameter and set a cap for the "trip_destination" model to ensure model identifiability.
• Modified the test_simple_simulate function in activitysim/estimation/test/test_larch_estimation.py to also use the "SLSQP" method for the "trip_destination" model. The SLSQP algorithm obeys constraints on parameter values, while the BHHH algorithm does not, yielding very different answers. Which is "better" depends on the modeler's opinion on the validity of the constraints.

Version Handling:

• Updated the larch version check in activitysim/estimation/larch/__init__.py to handle development versions correctly.

Model Parameter Updates:

• Updated the log-likelihood values in test_loc_trip_destination_BHHH_loglike.csv and test_loc_trip_destination_SLSQP_loglike.csv to reflect new model results. [1] [2]
• Added new size specifications for the "trip_destination" model in test_loc_trip_destination_SLSQP_size_spec.csv.

CSV File Updates:

• Updated parameter values in test_location_model_trip_destination_BHHH_None_.csv to reflect new model results.
• Added a new CSV file test_location_model_trip_destination_SLSQP_None_.csv with updated parameter values for the "trip_destination" model.