(it's #ifdef'ed out in aes_eax_aesni_test.{c,h} so test doesn't compile)

PiperOrigin-RevId: 214917122
GitOrigin-RevId: 434d8966fbf9c66c632c375c7fe8cc8e015aa48a

Our tests are linked with gunit_main (or gunit_main_no_google3), which means that main is already defined in another translation unit. Having two main functions violates ODR and hence can result in undefined behavior.

There is no need to fear losing tests. If a build system tries to run a test and there is no main function, then it will fail. Hence, i cannot see a way for this to stop testing a test.

PiperOrigin-RevId: 213593596
GitOrigin-RevId: 1fd8b2dcc18d24901d77170082587557cc5303e3

PiperOrigin-RevId: 206805592
GitOrigin-RevId: d2f6d3fd28e53ac323650486ca74754eac9d4417

PiperOrigin-RevId: 203378951
GitOrigin-RevId: 62fd8b1b82187a7758e7bbf66de363288305e3d2

(based on cl/196889713 and cl/197018177)

PiperOrigin-RevId: 202529544
GitOrigin-RevId: f290535f5a57642982d30c43b113cdae6fd7a385

library.

Things not done:
Things to check:
- The code does not use EVP. There are plans to just add code in assembly.
  Hence this is kind of backup code for the case where AES instructions are
  not available. Hence it is unclear whether using that horrible EVP interface
  makes sense.
- depot/google3/security/util/crypt/ contains a version of AES_EAX using
  EVP. The main problem with this implementation is that it is not thread safe.
- The code needs unaligned big-endian load and store. I have not been able
  to find a function that achieves this efficiently. Compilers can sometimes
  (but not always) optimize the load or stores.

PiperOrigin-RevId: 201177778
GitOrigin-RevId: 2601a0d6eb42804747718f283761774922cdf8f5