Handle jemalloc calls after TSD teardown

This only changes behavior for the tsd_generic path, where
pthread_getspecific() can report no TSD after the pthread key destructor
has finished.  Other TSD backends keep tsd_teardown_done() as a constant
false, so the added checks compile out there.

Avoid recreating TSD for late deallocations.  Preserve existing
reincarnation behavior for late allocations and nonzero reallocations.

Add Linux CI coverage for force_tls=0, with and without --enable-debug,
to exercise the Android-equivalent generic TSD path.

.github/workflows/linux-ci.yml

@@ -508,6 +508,16 @@ jobs:
               CXX: g++
               CONFIGURE_FLAGS: "--enable-debug --enable-experimental-smallocx --enable-stats --enable-prof"
               EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
+          - env:
+              CC: gcc
+              CXX: g++
+              CONFIGURE_FLAGS: force_tls=0
+              EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
+          - env:
+              CC: gcc
+              CXX: g++
+              CONFIGURE_FLAGS: "force_tls=0 --enable-debug"
+              EXTRA_CFLAGS: "-Werror -Wno-array-bounds"
           - env:
               CC: gcc
               CXX: g++