Clean up some code

- Use good ELF technique in cosmo_dlopen()
- Make strerror() conform more to other libc impls
- Introduce __clear_cache() and use it in cosmo_dlopen()
- Remove libc/fmt/fmt.h header (trying to kill off LIBC_FMT)

libc/nt/kernel32/FormatMessageA.S

@@ -0,0 +1,18 @@
+#include "libc/nt/codegen.h"
+.imp	kernel32,__imp_FormatMessageA,FormatMessageA
+
+	.text.windows
+        .ftrace1
+FormatMessageA:
+        .ftrace2
+#ifdef __x86_64__
+	push	%rbp
+	mov	%rsp,%rbp
+	mov	__imp_FormatMessageA(%rip),%rax
+	jmp	__sysv2nt8
+#elif defined(__aarch64__)
+	mov	x0,#0
+	ret
+#endif
+	.endfn	FormatMessageA,globl
+	.previous

libc/nt/master.sh

@@ -87,6 +87,7 @@ imp	'FindNextVolume'					FindNextVolumeW						kernel32	3
 imp	'FindVolumeClose'					FindVolumeClose						kernel32	1
 imp	'FlushConsoleInputBuffer'				FlushConsoleInputBuffer					kernel32	1
 imp	'FormatMessage'						FormatMessageW						kernel32	7
+imp	'FormatMessageA'					FormatMessageA						kernel32	7
 imp	'FreeConsole'						FreeConsole						kernel32	0
 imp	'FreeEnvironmentStrings'				FreeEnvironmentStringsW					kernel32	1
 imp	'FreeLibrary'						FreeLibrary						kernel32	1

libc/nt/process.h

@@ -51,6 +51,9 @@ void SetLastError(uint32_t dwErrCode);
 uint32_t FormatMessage(uint32_t dwFlags, const void *lpSource,
                        uint32_t dwMessageId, uint32_t dwLanguageId,
                        char16_t *lpBuffer, uint32_t nSize, va_list *Arguments);
+uint32_t FormatMessageA(uint32_t dwFlags, const void *lpSource,
+                        uint32_t dwMessageId, uint32_t dwLanguageId,
+                        char *lpBuffer, uint32_t nSize, va_list *Arguments);
 int64_t OpenProcess(uint32_t dwDesiredAccess, bool32 bInheritHandle,
                     uint32_t dwProcessId);
 uint32_t GetCurrentProcessId(void); /* %gs:0x40 */