staticdata: Remove new_ext_cis list

[topolarity]

As a root for compilation, this list is pretty much redundant with the external CI's that we will discover along the forward call edges in inference. As a filter for code emission, it is mostly redundant with specsigflags & 0b100, which is the only thing jl_emit_native respects as a filter right now anyway.

The other purpose of this list was to track what needs to be inserted into cache / methods loaded from another pkgimage (i.e. CI's for external methods), which is now tracked by a new specsigflags & 0b1000 flag.

This means that the list of "newly-inferred" CI's / MI's (as defined by JL_MI_FLAGS_MASK_PRECOMPILED / jl_push_newly_inferred / jl_tag_newly_inferred_enable) is purely a "best-effort" list - the runtime itself no longer depends on it.

TODO: The last remaining use for this list is to track which MI/CI's for external methods were triggered via precompile(...). Otherwise these will not make it into the queue.

[topolarity]

cc @timholy, to make sure this best-effort definition for "newly-inferred" / precompile-tagged is OK
(if not, let's try to find a precise definition)

[topolarity]

Oof. This does affect pkgimages significantly after all.

Compiling with JULIA_CPU_TARGET="generic;sandybridge,-xsaveopt,clone_all;haswell,-rdrnd,base(1);x86-64-v4,-rdrnd,base(1)", it looks like Pkg's image shrunk by about 10 MB (74 MB -> 65 MB) and basecompiler.so is down about 20 MB (78 MB -> 57 MB). A number of precompilation tests also got affected.

edit: Ah, I think I see the problem. precompile(...) statements for external methods are being dropped, since they are not discovered by this walk any more:

julia/src/precompile_utils.c

Lines 313 to 337 in 58b5fb8

	if (!suppress_precompile) {
	size_t i, n = jl_array_nrows(worklist);
	for (i = 0; i < n; i++) {
	jl_module_t *mod = (jl_module_t*)jl_array_ptr_ref(worklist, i);
	assert(jl_is_module(mod));
	foreach_mtable_in_module(mod, precompile_enq_all_specializations_, m);
	}
	n = jl_array_nrows(extext_methods);
	for (i = 0; i < n; i++) {
	jl_method_t *method = (jl_method_t*)jl_array_ptr_ref(extext_methods, i);
	assert(jl_is_method(method));
	jl_value_t *specializations = jl_atomic_load_relaxed(&method->specializations);
	if (!jl_is_svec(specializations)) {
	precompile_enq_specialization_((jl_method_instance_t*)specializations, m);
	}
	else {
	size_t j, l = jl_svec_len(specializations);
	for (j = 0; j < l; j++) {
	jl_value_t *mi = jl_svecref(specializations, j);
	if (mi != jl_nothing)
	precompile_enq_specialization_((jl_method_instance_t*)mi, m);
	}
	}
	}
	}

[vchuravy]

Yeah it's original purpose was to describe the root set of external methods who have new CI for a package image. This includes CI from a alternative abstract interpreter