- added constructor dictionary(const dictionary*) that also handles NULL
  pointers and makes it convenient to construct from a possibly nonexistent
  sub-dictionary:
  eg,
      dictionary dict2(dict1.subDictPtr("someDict"));

- make some of the turbulence Coeffs sub-dictionary optional.
  Their contents are all 'lookupOrAddDefault' anyhow.

- in turbulentMixingLength BCs, skip namespace qualifier in template
  (eg, <RASModel> vs. <compressible::RASModel>)

- change comments from 'turbulenceProperties' to RASProperties/LESProperties

- consistency between compressible/incompressible - no separate file for
  'New' selector etc

- consistency in accessing the model coefficients.
  Use method coeffDict() for const access.
  Use protected data member coeffDict_ for read/write access.

- document model coefficients in etc/constant/RASProperties.
  Need the same for LESProperties before we can prune these from the
  tutorials.

- added rules for gcc-4.3 in preparation for switching to gcc-4.4

- added 'clean' target for wmake/src/Makefile
  It is probably reasonable to separate the binaries from the rules (TODO)

- added obj surfaceWriter
- don't bother with prepended field names on obj and stl surfaceWriter.
  It fills the disk and suggests differences where there are none.

- put back debug switch (even although it is never used)
- remove stringsGlobals.C, since it is now never used

- string doesn't use any debug info, remove it
- restructured the macros to add in NoDebug macro versions to className,
  typeInfo.  Might be helpful with tackling the globals bootstrapping issue.
- HashTableName + StaticHashTableName - avoid lookup of debug switch when
  FULLDEBUG is not defined

- Problem: changing between OpenFOAM versions with different
  WM_PROJECT_INST_DIR paths would leave the old paths in the env vars.

- TODO: etc/cshrc is rougned out, but needs work on the corresponding
  if-statement

- The new cmake seems to be even smarter. If we try to use env variables in
  the .cmake files (eg, with And's sed fixup), they are rewritten in a
  subsequent re-make and take much longer).  Adjusting the compiler values
  themselves causes a full remake. On the positive side, disabling the rpath
  seems to mean we may not need any of this.

  In summary - yuck!

- set CMAKE_INSTALL_PREFIX:PATH now
- allow separation between -config/-make/-install stages
- update for cmake 2.6

  - avoid mkdir except for a few directories
  - grouped settings and added comments
  - minor bugfix: $FOAM_INST_DIR/linux64/bin was in path (pre OpenFOAM-1.5)
  - local (non-exported) variables in lowercase