liblzma: Fix building with NVHPC (NVIDIA HPC SDK).

NVHPC compiler has several issues that make it impossible to
build liblzma:
  - the compiler cannot handle unions that contain pointers that
    are not the first members;
  - the compiler cannot handle the assembler code in range_decoder.h
    (LZMA_RANGE_DECODER_CONFIG has to be set to zero);
  - the compiler fails to produce valid code for delta_decode if the
    vectorization is enabled, which results in failed tests.

This introduces NVHPC-specific workarounds that address the issues.

(cherry picked from commit 096bc0e3f8)

src/liblzma/common/string_conversion.c

@@ -217,12 +217,14 @@ typedef struct {
 	uint16_t offset;
 
 	union {
+// NVHPC has problems with unions that contain pointers that are not the first
+// members
+		const name_value_map *map;
+
 		struct {
 			uint32_t min;
 			uint32_t max;
 		} range;
-
-		const name_value_map *map;
 	} u;
 } option_map;
 

src/liblzma/delta/delta_decoder.c

@@ -25,6 +25,9 @@ decode_buffer(lzma_delta_coder *coder, uint8_t *buffer, size_t size)
 }
 
 
+#ifdef __NVCOMPILER
+#	pragma routine novector
+#endif
 static lzma_ret
 delta_decode(void *coder_ptr, const lzma_allocator *allocator,
 		const uint8_t *restrict in, size_t *restrict in_pos,

src/liblzma/rangecoder/range_decoder.h

@@ -45,6 +45,7 @@
 //     and different processors. Overall 0x1F0 seems to be the best choice.
 #ifndef LZMA_RANGE_DECODER_CONFIG
 #	if defined(__x86_64__) && !defined(__ILP32__) \
+			&& !defined(__NVCOMPILER) \
 			&& (defined(__GNUC__) || defined(__clang__))
 #		define LZMA_RANGE_DECODER_CONFIG 0x1F0
 #	else