Mostly revert r286392 (Roll BoringSSL.)

Mostly revert r286392 (Roll BoringSSL.)

There was an android failure which I am unsure if was due to BoringSSL roll:
   http://build.chromium.org/p/chromium.linux/builders/Android%20Tests/builds/14894/steps/androidwebview_instrumentation_tests/logs/stdio

I don't have time to investigate tonight, so backing out the majority of the BoringSSL roll.

The difference from a full rollback to 7e3305eebd7fb06d57e7f25b3bbf9c10d526f7d5  is that these two changelists have been kept (since I landed code that depends on RSA_check_key() already):

------------------

commit 05b7377065bd0acd705314a6411b734b67d0774e
Author: Adam Langley <agl@chromium.org>;
Date:   Fri Jul 25 12:03:51 2014 -0700

    Add RSA_check_key function.

    This is function that is available in OpenSSL too. Here it only returns
    zero or one and doesn't do expensive primality checks on p and q.

    https://code.google.com/p/chromium/issues/detail?id=396250

    Change-Id: I7a173da26e06440dbb595fb717e3a620edf23576
    Reviewed-on: https://boringssl-review.googlesource.com/1334
    Reviewed-by: Adam Langley <agl@google.com>;

commit 93a3dcd57262f5bbd84adaa5565abb14f95d79ae
Author: Adam Langley <agl@chromium.org>;
Date:   Fri Jul 25 15:40:44 2014 -0700

    Add AES Key Wrap mode.

    This is needed in order to support Web Crypto.

    https://code.google.com/p/chromium/issues/detail?id=396407

    Change-Id: I900d8cad2716c2e3341eeae153659502326c9173
    Reviewed-on: https://boringssl-review.googlesource.com/1335
    Reviewed-by: Adam Langley <agl@google.com>;

------------------

TBR=agl@chromium.org

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=286419

[eroman, 2014-07-30 06:05:32]

Committed patchset #1 manually as r286419 (presubmit successful).

[agl, 2014-07-30 17:52:19]

LGTM, although it looks unrelated:

http://build.chromium.org/p/chromium.linux/builders/Android%20Tests/

androidwebview_instrumentation_tests is still failing often.

[eroman, 2014-07-30 18:08:24]

Agreed, I don't think BoringSSL roll was the cause of the flaking android tests.

Here is my brain dump if you are curious:

  - The very first failure of android webviewtests yesterday was at r286374,
with just that changelist in the window
--http://build.chromium.org/p/chromium.linux/builders/Android%20Tests%20%28dbg%29/builds/21881

This breakage preceded the BoringSSL roll (assuming the revision ranges reported
by the bot are accurate).

Moreover once I mostly reverted BoringSSL the flake persisted (the two
changelists I kept seem highly improbably at breaking things, so I am fairly
confident this couldn't be it).