* add crl.get_revoked_certificate method

* lexicographic is the best ographic

* rename

And not expose an unprefixed name to anyone who dlopens us.

* set an OPENSSL_API_COMPAT level

this helps prevent adding deprecated functions and will let us see what
we need to/can prune in the distant future when we support only 1.1.0+

* raise the api compat to 1.0.1 (which doesn't matter but is less confusing)

In #4220, I switched BIO_new to take a const pointer (true in 1.1.0, but
not 1.0.2) on grounds that:

    This also folds in the const bits from 1.1.0, on the assumption that,
    now that the function pointer check is gone, it will just cause cffi to
    generate more conservative pointer types that work for 1.0.2 as well.

But I got this backwards. If the bindings say BIO_METHOD*, cffi will
pass a BIO_METHOD* to BIO_new, which works in both OpenSSL versions. If
it says const BIO_METHOD*, cffi will pass const BIO_METHD* to BIO_new,
which does not work in 1.0.2.

(Although cryptography.io's build ignores all these warnings anyway, so
it's kind of moot.)

These are required by the Tribler project.

* Perform an OPENSSL_cleanup before checking the heap in our memleak tests

* Make this binding conditional

* typo

* need to put this call before we reset the function ptrs

* libre 2.7.3 compatibility

* add a changelog

* actually build against 2.7.3

* Remove defines for openssl 1.1.0 pre

* Update bio.py

* Update dh.py

* Update dsa.py

* Update rsa.py

* Update x509_vfy.py

* Compress branches

* Fixed build errors on HP-UX.

* PEP 8 style fix.

* No return for void function.

* PEP 8 style fix, take 2.

This removes:

- ERR_get_state which really shouldn't be public API.

- A bunch of functions that are really mostly useful within the library
  to add new errors. NB: I say mostly because they are also useful when
  trying to register a new error library, as osrandom does, but osrandom
  is written in C. Python code is more likely to be consuming errors.

- All function codes but EVP_F_EVP_ENCRYPTFINAL_EX because tests still
  reference it. Per PR #3609, function codes are kind of unstable. This
  finishes that up and cleans up the bindings.

- The "line" versions of querying the error queue, just because no one
  seems to be using them and there's a lot.

- Error-printing functions, which make less sense in Python since you'd
  probably wrap in an exception.

Error codes probably could also do with cleaning, but I've left them
alone for now.