Improve handling for tryResolve and errors to handle a bad init=

[mppf]

In the event that init= indicates that a type is not copyable
(currently by including a compilerError in the body), the compiler
should raise an error if the copy is used but not if the copies are all
removed. Since some copies are actually removed during callDestructors,
this PR extends the work of PR #8577 with FLAG_ERRONEOUS_INITCOPY.

Changes are:

• factor some ast-location code into compiler/util/astlocs.cpp from
  baseAST.cpp and misc.cpp
• add a userInstantiationPointLoc to TypeSymbol and FnSymbol - used
  strictly for error messages - and add getUserInstantiationPoint to
  astlocs.cpp to help with this
• change from FLAG_ERRONEOUS_AUTOCOPY / FLAG_ERRONEOUS_INITCOPY to
  just FLAG_ERRONEOUS_COPY (it's simpler and it can also apply to
  array's chpl_unref)
• in initializerRules, allow type queries on this (for partial
  instantiations) so that the range constructor can work correctly when
  called on an instantiation
• remove defaults from resolveUninsertedCall to reduce confusion
• adjust resolveUninsertedCall to return the result of tryResolve if
  that is the strategy used (so that tryResolve can return NULL, e.g.)
• check for an error saved in tryResolveErrors in case it's the 2nd time
  an init= (say) was tryResolved. (Otherwise, the error would only be
  encountered the first time, and then it would be saved in the map, but
  the call would resolve to the function containing the error - leading
  the tryResolve to succeed).
• add fixInstantiationPointAndTryResolveBody and call it from several
  type-directed functions (like resolveAutoCopyEtc, findAssignFn, etc).
  This detects compilerErrors in the called function and sets the
  instantiation point appropriately.
• factor most of the code in preFold for PRIM_IS_COPYABLE etc into
  separate functions
• set instantiationPoint on the TypeSymbol for a constructed tuple type
• make handleError accept an astlocT and use that instead of the hacky
  createASTforLineNumber - and while there get one more handleError
  variant to call vhandleError instead of repeating the same code
• add a compilerError for initCopy for arrays to check that the type is
  assignable
• adjust range.init to work with (partial) instantiations (so that
  isDefaultInitializable will work correctly on it)
• add tests of missing init= and arrays of types missing =
• update several chameneous tests to use nilable types to work around
  issue cannot return array of non-nilable owned #14896

Reviewed by @benharsh - thanks!

• full local futures testing

[mppf]

Just noting that this PR is connected to issue #8065 which discusses how users should indicate a type is not copyable.