fix alignment issue for bele tests

[lemire]

The test is poorly written and on some system, it may lead to unaligned access when running the tests (armv7). Thanks to @clausecker for proposing a fix.

Fixes #341

[clausecker]

LGTM, though you should be able to just do

const char32_t utf32 = { 0x40, 0xa7, 0x2208, 0x01d4aa };

which should do the right thing on both byte orders. For UTF-16, there's probably a similar approach, but I'm not exactly sure what it would be.

[lemire]

@clausecker Thanks. I think that this testing code should be redone in a cleaner fashion.

Let me issue a release now.