Wrap boost::asio::async_write to ensure lifetime of buffers

[wezrule]

Introduce a new class nano::shared_const_buffer which @cryptocode found an example for here: https://www.boost.org/doc/libs/1_69_0/doc/html/boost_asio/example/cpp11/buffers/reference_counted.cpp This should alleviate any buffer memory not being captured or having sufficient lifetime during the boost::asio::async_write call. When building travis I check if boost::asio::async_write is called anywhere other than the new file which wraps it nano/lib/asio.hpp and error out with a message if it is.

I think any minor performance hits we have by always using a shared_ptr are outweighed by keeping the program correct, as this has bitten us in the past and I found yet more places (nano::rpc_request_processor::try_reconnect_and_execute_request) where this wasn't done properly even though we actively try make sure it is.

[cryptocode]

For another PR, but would something similar work for reads? I.e. a shared_mutable_buffer class (using boost::asio::mutable_buffer)?

[wezrule]

The MutableBufferSequence has similar requirements to the ConstBufferSequence so I think it should be fine to add something similar for async_read
https://www.boost.org/doc/libs/1_67_0/doc/html/boost_asio/reference/MutableBufferSequence.html