fmt/pl: optimize pl_strstr, pl_strchr and pl_strrchr

[sreimers]

Replace byte-by-byte scan with optimized memchr first character matching and confirm with memcmp.

[alfredh]

unrelated comment, this hand-crafted function can probably also be replaced
by system functions:

#ifndef HAVE_STRINGS_H
static int casecmp(const struct pl *pl, const char *str)
{
	size_t i = 0;

#define LOWER(d) ((d) | 0x20202020)
	const uint32_t *p1 = (uint32_t *)pl->p;
	const uint32_t *p2 = (uint32_t *)str;
	const size_t len = pl->l & ~0x3;

	/* Skip any unaligned pointers */
	if (((size_t)pl->p) & (sizeof(void *) - 1))
		goto next;
	if (((size_t)str) & (sizeof(void *) - 1))
		goto next;

	/* Compare word-wise */
	for (; i<len; i+=4) {
		if (LOWER(*p1++) != LOWER(*p2++))
			return EINVAL;
	}

 next:
	/* Compare byte-wise */
	for (; i<pl->l; i++) {
		if (tolower(pl->p[i]) != tolower(str[i]))
			return EINVAL;
	}

	return 0;
}
#endif