Backport #45727 and a patch to make it work (#45938)

* Bug fix for `max_collect_interval` computation (#45727)

Currently constrained to `totalmem / ncores / 2` for `_P64` which
results in a very short collect interval when you're running with
a smaller number of threads on a machine with many cores.

Changes this to `totalmem / nthreads / 2` which, for two of our
tests, resulted in 40% and 60% runtime reduction (!!) as well as
GC time reduction from 46% to 10% and 64% to 11%.

* Move GC init after threading init

To allow use of `jl_n_threads` in GC initialization.

src/gc.c

@@ -3410,12 +3410,12 @@ void jl_gc_init(void)
     gc_num.allocd = 0;
 
 #ifdef _P64
-    // on a big memory machine, set max_collect_interval to totalmem / ncores / 2
+    // on a big memory machine, set max_collect_interval to totalmem / nthreads / 2
     uint64_t total_mem = uv_get_total_memory();
     uint64_t constrained_mem = uv_get_constrained_memory();
     if (constrained_mem > 0 && constrained_mem < total_mem)
         total_mem = constrained_mem;
-    size_t maxmem = total_mem / jl_cpu_threads() / 2;
+    size_t maxmem = total_mem / jl_n_threads / 2;
     if (maxmem > max_collect_interval)
         max_collect_interval = maxmem;
 #endif

src/init.c

@@ -686,10 +686,10 @@ JL_DLLEXPORT void julia_init(JL_IMAGE_SEARCH rel)
     }
 
     jl_init_runtime_ccall();
-    jl_gc_init();
     jl_init_tasks();
     jl_init_threading();
 
+    jl_gc_init();
     jl_ptls_t ptls = jl_init_threadtls(0);
     // warning: this changes `jl_current_task`, so be careful not to call that from this function
     jl_task_t *ct = jl_init_root_task(ptls, stack_lo, stack_hi);