Update boost on coverage pipeline [skip-ci]

Build fedora, boost version 1.67

chain ID is now handled at runtime and is the hash of the genesis state; also the genesis state is included at the start of blocks.log

Add configuration arguments [skip-ci]

This reverts commit 0eda814c.

chain_id_type should be its own type (without a public default constructor) rather than just a typedef of fc::sha256 #3415

Special exceptions had to be made to `fc::variant::as()const` and
chain_plugin and net_plugin to access the private default constructor.
This is an ugly hack but it works for now. Eventually I need to see if
there is a better solution.

Update boost to 1.67.0 in builder [skip-ci]

Remove snapshot submodule.

Merging master into branch eosio_build_fedora

bugifxes: delegated bandwidth #3366, buyram from eos #3414

Note that for existing checkouts, a copy remains in $GIT_DIR/modules so
checkout of previous revisions does not require fetching from a remote
repository.  New checkouts will not have a copy unless an older revision
is checked out, followed by git submodule update --init --recursive.

system contract undelegatebw bugfix: don't delete row if ram_bytes != 0, small change in resource_limits to make debuging less confusing #3366

Also the initial public key of the 'eosio' account can now be specified
in CMake rather than specifying the chain ID.

Don't use eosio.token for txn_test_gen_plugin; make separate account

Change producer deactivation logic

bios-boot-tutorial: command line, increase producer stake

Don't use the eosio.token account for txn_test_gen_plugin usage; this frequently clashes will a real network. Create another account where we upload eosio.token for txn_test_gen_plugin usage