Make minor improvements

- Work towards simplifying ape.S startup process
- Rewrote ar because it took minutes to build cosmopolitan.a

libc/nexgen32e/crc32init.S

@@ -59,7 +59,7 @@ crc32init:
 	pand	%xmm0,%xmm3
 	pxor	%xmm4,%xmm3
 	movdqa	%xmm3,%xmm4
-	loop	2b
+	.loop	2b
 	movdqu	%xmm3,(%rdi)
 	add	$16,%rdi
 	paddd	%xmm2,%xmm1

libc/nexgen32e/imapxlatab.S

@@ -35,7 +35,7 @@ imapxlatab:
 	.align	8
 1:	stosq
 	add	%rdx,%rax
-	loop	1b
+	.loop	1b
 	.leafepilogue
 	.endfn	imapxlatab,globl,hidden
 	.source	__FILE__

libc/nexgen32e/kbase36.S

@@ -35,13 +35,13 @@ kBase36:.zero	256
 	pushpop	10,%rcx
 0:	inc	%eax
 	stosb
-	loop	0b
+	.loop	0b
 	add	$'A-1-'9,%rdi
 	pushpop	'Z+1-'A,%rcx
 0:	inc	%eax
 	mov	%al,0x20(%rdi)
 	stosb
-	loop	0b
+	.loop	0b
 	add	$255-'Z,%rdi
 	.init.end 300,_init_kBase36
 	.source	__FILE__

libc/nexgen32e/ktolower.S

@@ -51,7 +51,7 @@ kToLower16:
 	mov	$256,%ecx
 0:	lodsb
 	stosw
-	loop	0b
+	.loop	0b
 	pop	%rsi
 	.init.end 300,_init_kToLower
 

libc/nexgen32e/ktoupper.S

@@ -30,6 +30,6 @@ kToUpper:
 	call	imapxlatab
 	pushpop	'z-'a,%rcx
 0:	subb	$0x20,(%r8,%rcx)
-	loop	0b
+	.loop	0b
 	.init.end 300,_init_kToUpper
 	.source	__FILE__

libc/nexgen32e/memjmpinit.S

@@ -36,7 +36,7 @@ memjmpinit:
 	lodsb
 	add	%rdx,%rax
 	stosq
-	loop	0b
+	.loop	0b
 	xor	%eax,%eax
 	testb	X86_HAVE(ERMS)+kCpuids(%rip)
 	setnz	%al

libc/nexgen32e/rldecode.S

@@ -34,7 +34,7 @@ rldecode:
 	lodsb
 	jrcxz	2f
 1:	stosb
-	loop	1b
+	.loop	1b
 	jmp	0b
 2:	.leafepilogue
 	.endfn	rldecode,globl