update modcc to latest modparser trunk

* update the out of date version of modparser was added to the repository
* fix some warnings about unused static functions in a modcc test header
* move modcc/src path contents to modcc, because there was little point having
  the additional sub directory.

Moved over modcc unit tests.

Can't std::sort on forward iterators.

But GNU libstc++ tries to anyway, hilarious segfaults
ensue.

* Don't try and sort a sentinel-terminated range.
* Instead provide a `strict_view` that walks a range
  (if required) and presents a range with a proper
  end iterator.
* Tests, const range (but non-const iterator) implementations.

pid doc cleanup

generated mechanism headers depend on modcc

fixed modcc->hpp dependencies

move modparser from a submodule to the repository

* only the source has been moved over and the cmake rules updated
* the tests still have to be moved

move vector library from submodul to subdirectory

Cleanup

Cleanup

Fix for warnings with latest intel compiler.

Feature/sort synapse indices

Present an ordered sequence of compartment indices
to point processes (synapses) on creation in fvm_multicell.
Required for GPU implementations.

Also:
* Change target handle for fvm_multicell to refer to actual
  (rather than offset) mechanism index.
* Split out the range/sequence utilities that were accumulating
  in `util/range.hpp` into their own header `util/rangeutil.hpp`.
  Possibly `sequtil.hpp` is a better name…
* Add range/sequence-friendly `util::sort` and `util::sort_by`.
  The latter sorts based on comparing a derived key.
* Add `util::assign_by`. Assigns after applying a transformation.
* Fix `canonical_view` for non-const ranges/sequences.
* Unit tests for `assign`, `assign_by`, `sort`, `sort_by`.

Also updates modparser to point to the latest master.

Add script for analysis of MPI profile output.

Script collates region times from JSON-formatted profile
output, presenting raw times or quartiles in CSV table.

Bugfix/gcc 4.9.3 type test workaround

Void default argument SFINAE test for presence of
type definition within in a class fails in gcc 4.9.3.

This commit uses a different SFINAE approach to work around
the compiler bug.

Feature/fvm multicell