Fixes to symmetric decryption

[einarmo]

Remove padding when decrypting symmetric messages.

Also fixes a pair of minor issues:

• We stored the session ID before actually validating that we got a valid session. This stuff can still be improved, but this at least fixes an obvious hole.
• 2^13 is 8192, not 8196, this is also what the standard uses.

Should fix #373

[DOhlsson]

I tried running with the code in this branch but got the error I posted about here

[DOhlsson]

I tried running with the code in this branch but got the error I posted about here

That was an unrelated issue. This PR solves the actual problem I was having.

[svanharmelen]

I've spend a day or two debugging what felt like random error messages when making browse requests to a reasonably big (so with a few nodes with many, many children) OPC server. Here are two examples:

"Failed to fill whole buffer":

"String exceeds decoding limits" (while the actual strings are all less then 30 bytes long):

I noticed that the issue only occurred when the message was split into multiple chunks, so I've had a closer look at all the chunker logic. By the time I was done reading the specifications and a few HEX dumps of all the bytes coming in and being reconstructed to a singe message, I too came to the conclusion that the padding info was not being removed.

Luckily I decided to check for open issues and PRs once more (now I knew what to look for) before creating a PR myself and found this PR 😏

@locka99 it seems you are pretty busy lately, but this is actually a pretty nasty bug. So if/when you have a minute it would be great if you can have a quick look and/or merge this PR. Also "hello again 👋🏻" as I've been away from this project for quite a while, but I'm back at it for a new project which should keep me busy for years to come 😉