Add target-byteorder for cases where endian in target triple is what matters (#107915)

I came across the subtly when setting up lit for z/OS and running it on
a Linux on Power machine. Linux on Power is little endian. This was
resulting in all of these tests being run even though the target triple
was z/OS which is big endian. The lit should really be checking if the
target is little endian not the host. The previous way didn't handle
cross compilation while running lit.

Author: perry-ca

llvm/test/CodeGen/Generic/allow-check.ll

@@ -1,5 +1,5 @@
 ; Avoid `!DL->isLittleEndian() && !CLI->enableBigEndian()` missmatch on PPC64BE.
-; REQUIRES: host-byteorder-little-endian
+; REQUIRES: target-byteorder-little-endian
 
 ; -global-isel=1 is unsupported.
 ; XFAIL: target=loongarch{{.*}}

llvm/test/lit.cfg.py

@@ -364,6 +364,14 @@ def version_int(ver):
     config.available_features.add("llvm-64-bits")
 
 config.available_features.add("host-byteorder-" + sys.byteorder + "-endian")
+if config.target_triple:
+    if re.match(
+        r"(aarch64_be|arc|armeb|bpfeb|lanai|m68k|mips|mips64|powerpc|powerpc64|sparc|sparcv9|s390x|s390|tce|thumbeb)-.*",
+        config.target_triple,
+    ):
+        config.available_features.add("target-byteorder-big-endian")
+    else:
+        config.available_features.add("target-byteorder-little-endian")
 
 if sys.platform in ["win32"]:
     # ExecutionEngine, no weak symbols in COFF.