Don't skip calling the cleanup hook in EVP_PKEY_CTX when the copy hook fails

This was only reachable from malloc failure. Way back in
https://boringssl-review.googlesource.com/c/boringssl/+/13830 we
imported a change from BoringSSL to stop calling cleanup when
EVP_PKEY_CTX_dup failed. This supposedly fixed a crash (I didn't check
if it did at the time) but instead caused different malloc failures to
leak memory.

Now that all our cleanup functions are just a call to C++ Delete, and we
use C++ scopers for all the fields, the EVP_PKEY_CTX state is in a
consistent enough state to delete all the time anyway.

While I'm here, I've fixed the cleanup functions to all be more
consistent. Some of them reset ctx->data and some of them didn't. It
doesn't matter if we do because the object will be destroyed immediately
afterwards, so I standardized on the shorter pattern.

To review this, check that a failure in the copy hook never leaves the
EVP_PKEY_CTX in a state where the cleanup hook will not safely clean
this up. Note that bssl::Delete, like C++ delete, gracefully handles
nullptr.

Change-Id: I874b953efbef179f1387c0a2b2e0320e6c3ff2e6
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/93627
Commit-Queue: David Benjamin <davidben@google.com>
Reviewed-by: Rudolf Polzer <rpolzer@google.com>
Auto-Submit: David Benjamin <davidben@google.com>