Fix setData

[paciorek]

This fixes issues #1324 and #1326 by:

• only overwriting values set by setData with setInits if the value from setData is NA.
• warning users if they try to overwrite a value set by setData in setInits and the setInits value is different from the setData value and not NA.
• adding an includeRHSonly flag to isDeterm with default FALSE and then updating some uses of isDeterm in our codebase to set includeRHSonly to TRUE.
• fixing #1326 by indexing into varValue (when checking for deterministic data) and isDataVarValue (when setting data flags) using eval with the scalar node names (similar to how isDataFromGraphID works). This avoids setting the data flag to TRUE for deterministic and RHSonly elements of a variable that has data elements. Note that positions in isDataEnv$x can still be TRUE for positions that do not correspond to any nodes (e.g., if x[1] is not defined at all in the model, but the user provides a non-NA value for x[1] in setData).

Note that I have not yet added various tests, in particular the CR-type example where there are undefined nodes in a variable, such as y[i, 3:6] defined but y[i, 1:2] undefined. I will also review the changes again carefully.

[paciorek]

Per discussion in NCT issue 457, this approach won't work because it prevents use of setInits to generally change values. See next PR that avoids the RHSonly as data issue and simply addresses determ nodes as data.