Explicitly adding channel bindings on the client

In the current API I see no way to add channel bindings to an `Ntlm` context when authenticating a client. (Motivation: AD can be configured to refuse authentication from an LDAP client doing an NTLM bind over a TLS connection if the channel bindings AvP is not supplied.)

Adding a method to `Ntlm` which sets the `channel_bindings` field to the supplied `ChannelBindings` struct does the right thing protocol-wise, i.e., the channel bindings AvP is correctly calculated and added to the AUTHENTICATE message. Would a PR with this kind of addition be appropriate, or is there another preferred way to add this functionality to the library?

Start by locating the Ntlm client context and its channel_bindings field, then trace how ChannelBindings is used when constructing the AUTHENTICATE message. Check existing Ntlm API tests or authentication tests for the appropriate coverage. Done means clients can supply a ChannelBindings value and the resulting message includes the calculated channel bindings AvP.

Written by the indexing model from the issue text.