Add support for minimal thread local data

[hkaiser]

If HPX_WITH_THREAD_LOCAL_DATA is not defined, we now support storing a single size_t as thread local data

This also addresses part of #2371

@dhollman this should enable what you need by default and with reasonably small overheads

[dhollman]

@hkaiser Shouldn't this extension also be made to work with hpx::this_thread::get/set_thread_data() ?

[hkaiser]

Shouldn't this extension also be made to work with hpx::this_thread::get/set_thread_data()

@dhollman These functions should work as well (see

hpx/hpx/runtime/threads/thread.hpp

Lines 216 to 217 in 6d5b11e

	HPX_API_EXPORT std::size_t get_thread_data();
	HPX_API_EXPORT std::size_t set_thread_data(std::size_t);

)