documentation: unify API of compile time block merging vs dynamic block connection

[wirew0rm]

At the moment we have different API for the different modes of "combining multiple nodes" with different behaviour and API:

	process_one(_simd)	process_bulk	buffer	settings & tags handling
merged_node	yes	no	eliminated*	not yet
connect static	yes	yes	yes*	yes
connect dyn	yes	yes	yes*	yes

* as mege node stores the original blocks and these contain their ports which then contain their buffers, the buffers are at the moment just unused, if i read the code correctly.

** at the moment the ports and buffers are set up when the node is constructed and the connections between the port buffers are set up when the graph is initialized by the scheduler. This means that dynamic nodes always have to pass the data though a buffer, which is not optimal and could be eliminated in a lot of cases

The different APIs look like this:

// merged_node
auto merged = merge<"scaled", "addend1">(scale<int, 2>(), adder<int>());

// dynamically connect_static ports
auto &src = flow_graph.make_node<Source<float>>({ { "n_samples_max", n_samples } });
auto &block1 = flow_graph.make_node<TestBlock<float>>();
flow_graph.connect<"out">(src).to<"in">(block1));

// connect dynamic ports (from prototype implementation, not yet implemented in core)
auto &adder = flow_graph.add_node(std::make_unique<multi_adder<double>>(n));
for (i = 0; i < n; i++) {
    auto &source = flow_graph.make_node<fixed_source<double>>();
    flow_graph.dynamic_connect(source, 0, adder, j);
}