Fix IndexError in `generate_eagle_track`

[ewquon]

Changes:

• Add call to Simulator._adjust_track_start_region() to calculate an adjusted Config.track_start_region as needed, prior to calling movmodel.get_starting_indices()
• Catch IndexError in movmodel.generate_eagle_track() at mean_cond = 2.0 / (1.0 / local_cond[1, 1] + 1.0 / local_cond) which arises due to an invalid (empty) local_cond at the domain boundary.
• Changed track_start_region in examples/example_jem.py to be a 1-km thick region for now; need better workaround or should update inputs for other examples
• Add option when instantiating the Simulator object to specify seed=### for reproducible results (appears to work as expected in parallel as well).

Adjusting the track start region appears to be a bandaid fix. E.g.,

• Without adjusting the start region: 80/1000 were cut short, after about 3 moves on average
• Shifting the edge of start region 5*dx from the boundary: 16/1000 were cut short, after about 8 moves on average
• Shifting the edge of start region 10*dx from the boundary: 2/1000 were cut short, after 11 moves

[ewquon]

@rimplesandhu, does burnin=10 make sense for this existing boundary catch? https://github.com/NREL/SSRS/blob/fbc6687372edb0f761983a2e55e2aa802a3c40d3/ssrs/movmodel.py#L246-L248

[rimplesandhu]

@rimplesandhu, does burnin=10 make sense for this existing boundary catch?

https://github.com/NREL/SSRS/blob/fbc6687372edb0f761983a2e55e2aa802a3c40d3/ssrs/movmodel.py#L246-L248 this issue is resolved in the updated master branch