Fixed other formulas for CompletingLatentPeriod to not scale by dt, given that resulting net flows are multiplied by dt

Fixed formula for CompletingLatentPeriod to not scale by dt, given that resulting net flows are multiplied by dt

Added comments at the top of the .c file to clarify that there is only one currently active model and that it must be run from a corresponding R harness, and to note the need to use ulimit -c

Added comments in the .c file to clarify that currently active model must be run from a corresponding R harness

Added support for the SEIRWithLogCParameter scenario, so as to use stochastic evolution and a more generous/accomodating r dispersion parameter (100 vs. 1000, thereby allowing a broader likelihood function

Added support for the SEIRWithLogCParameter scenario, improved indication of how to use the C interface from different models (for now, until separate modules supported), improved logging

Cleaned up logging to revert to more reasonable levels for debugging-related logging messages still of interest for genreal logging

Broadened the conditions of resampling to be done conceptually before the model is advanced (at the end of the previous observation, even for the initial timeepoint), as appears to be consistent with step 2b in Andrieu et al. p272

Added code to include the initial values of parameters and latent stated within the recorded values; it had previously been omitted -- something which further impeded interpretability, as initial values were uninitialized

Fixed key bug within  copy3DDoubleArray whereby one of the indexes was iterating over the count of elements for the other index

Fixed key bug within  copy3DDoubleArrayLooselyTyped whereby one of the indexes was iterating over the count of elements for the other index

Updated R code so that arrays of ints for timepoints and observations that should be arrays of double are now properly arrays of double

Modified limited span/scope of stateVariableByParticleTimeStateVariable by confining it to performSMCAndSampleParticularStateVariableTrajectory (the function that performs SMC and samples the trajectory).  This can be done because -- per Juxin Derivation of Calculation of p sub Theta of y1.pdf and equation 9 -- values of state variables are not required to compute p sub Theta of y1 through yT; rather, just the value of weights are.

For greater clarity, changed function name stateVariablesSampledByTimepointAndStateVariable to sampledStateVariablesByTimepointAndStateVariable

Modified to account for the fact that computeProbOfAllObservablesPSubThetaOfY1ThroughT does not require knowledge of state variable values, only values of weights

Modified to account for the fact that only unnormalized weights are used in Juxin Derivation of Calculation of p sub Theta of y1.pdf; had to pass around both unnormalized weights more broadly for this reason, and normalized weights when performing resampling and picking a trajectory

Modified to account for Juxin Derivation of Calculation of p sub Theta of y1.pdf, as of Juxin's not of January 15, 2018

Improved and more consistent logging levels, allowing meaningful high-level output even for levels 0 and 1

Misc debugging bug fixes and changes to constants, as well as additions of many diagnostic test points

Per Andrieu p272, fix to ensure that no resampling or ancestor array accumulation takes place for initial timepoint

Updated to use regular weights and probabilities rather than log versions thereof, removed Mohammad Obeidat code