Unify callbacks with block transfer and non-blocktransfer

[mlasch]

Currently for each operation there are two possible callbacks, either with or without block transfer. However it is not always known a priori whether a block transfer callback is needed.
For example a "normal" read lwm2m_dm_read can induce a block response which triggers callbacks without block information.
I suggest having a block information structure in every callback which keeps track on the block transfer and is NULL if the callback results from a non-block transfer. Something like:

typedef struct _block_information_
{
  int block_num;
  int block_size;
  bool block_more;
} block_information_t;

However this would break the current API, any thoughts?

[sbernard31]

About breaking the API, this discussion makes me feel that this is probably not a problem at least for master.

[tuve]

I think having a unifed api with an optional struct as described is much better than the current solution. The reason I didn't do something like that when I implemented block transfer was that I didn't wanted to break the current api, but moving forward I think this is the way to go.

[sbernard31]

Fix by #540.