Now happens when JS canvas want toggle log states

Later such specifier can be used to access object parts like axis.
For the moment is not relevant

Use no_root_json and no_h_json URL parameteres.
Automatically applied for webcanvas

Normally THttpServer includes current object version when 
reply on draw.htm requests. With "no_root_json" and "no_h_json" 
URL parameters one can disable this. Such functionality required 
for web canvas

Let retrieve latest canvas version even without explicit
Canvas::Modified() call

Fixes threading issues: multiple threads storing and restoring mutex state, the
stack being modified concurrently etc.

This is kosher because the locks were taken to forbid other threads from modifying state while this thread is modifying state.
But this thread will not modify any state while the user code is active; thus while that user code is active,
the mutex can be completely reset, and restored once the user code returns.

Reset() completely unlocks a locked mutex, returning a state object that allows a
subsequent call to Restore() to restore that state.

Useful for locking / unlocking cling when used in MT context.

When forComp is passed to Map, the date and time is omitted so one can easily compare
two version of the 'same' file produced at different time.

Also update the output format to be cleaner.

The pragmas 'GCC diagnostic push/pop' also have an effect on Clang,
so we can reuse the push/pop from GCC pragmas and only add the
extra warning to be suppressed for Clang (GCC does not know about
-Wconditional-uninitialized).

Reference: http://cdash.cern.ch/viewBuildError.php?type=1&buildid=390509

Suppressing -Wall is not enough to avoid -Wconditional-uninitialized
warnings from Vc with Clang, so we need to suppress this one explicitly.

The static_assert() in Vc/common/simdize.h:758 is triggering due to
some strange interaction between the pragmas to suppress the warnings
and other things (see commit a79d073e). Newer versions of GCC seem
unaffected by this problem, so we require a recent enough GCC from
now on to sidestep this issue.