backwards compatibility

All other functions should not have been exposed in the shared library but were: now fixed by mixing the visibility flag into CFLAGS.

updated how the evolution is stopped in terminate_evolution: it now just sets stardata->evolving to FALSE so everything stops.

cleaned up buffered_printf also

fixed buffer_empty_buffer and clear_printf_buffer to set the buffer_alloced to zero (otherwise buffered_printf gets confused)

remove deprecated and unused miniz

rebuilt the buffered_printf code to simplify it and prevent (hopefully) the crashes that Alex Kemp finds.

added tmpstore->error_buffer to deal with errors separately,
this is handled correctly in binary_grid2 (v2.1.3+) so it
behaves just like previously and SYSTEM_ERROR is detected
correctly.

deprecated the printf_stack to remove a load of unusued code

extended the API to use the error_buffer through binary_c_error_buffer()

fixed the APItest directory : especially the fortran code didn't
work with the latest gfortran, seems ok now (for now?)

sync latest perl

timers was broken unless on an intel-compatible CPU, now we use the standard timer functions where rdtsc ticks are not available

fixed formatting of size_t and other types

update mega/mebi bytes

fixed ints to Star_numbers where appropriate

fixed nucsyn abundances update to use previous timestep's abundances when appropriate - I hope this is working now, tests are running

also updated docs to include a section on coding style

cleaned up some format statements using the gnu_printf attribute

buffered_printf now uses a 1.2x multiplier to update the size of its buffer, instead of matching the string exactly

put newstar star_t structs in tmpstore to avoid multiple allocations in stellar_structure and stellar_evolution

more cleanup of old code to try to speed it up

updated core-omega coupling

fix bugs in version() which caused it to crash when stardata is NULL

fixed PRINTF to output to stdout when stardata is NULL

fixed post-AGB logging in log_every_timestep (it requires DISCS)

cleaned up for merge back to master

This involved fixing the configure script for the default
arch/build target, and fixing the va_list behaviour is as an
array on x86_64 and not on ARM. Taking a pointer to it fixed this.

Also fixed the make_data_objects scripts to take obcopy's opts as
an environment variable so they can be controlled by the configure
script.

fixed but with max_hewd_mass

the buffering printf function had a buffer overrun in it (the irony)
which was triggered especially by the new version function being long.
I have fixed this and increased the size of the string buffer so that the
version string fits... in future this may need further extension.

fixed the prevention of exiting on nan because of a new parameter that
has "nan" in its name. This no longer fails: the issue was that stardata
is NULL when new_system is setting up, it's *new_stardata that must be
set and tested.

Debug_expression is only now evaluated when stardata is non-NULL in 
debug_fprintf.

Cleaned up various debugging statements as well.

also updates to the disc code

Added ability to convert from a header (or general data) file to a binary object which can then be linked directly. This prevents gcc from accessing the file and using huge amounts of RAM at build time. It also fixes the nucsyn_extended_s_process table which would not compile under Intel's compiler.