Improve freezer DB efficiency with periodic restore points (#649)

* Draft of checkpoint freezer DB

* Fix bugs

* Adjust root iterators for checkpoint database

* Fix freezer state lookups with no slot hint

* Fix split comment

* Use "restore point" to refer to frozen states

* Resolve some FIXMEs

* Configurable slots per restore point

* Document new freezer DB functions

* Fix up StoreConfig

* Fix new test for merge

* Document SPRP default CLI flag, clarify tests

beacon_node/src/cli.rs

@@ -1,4 +1,5 @@
 use clap::{App, Arg, SubCommand};
+use store::StoreConfig;
 
 pub fn cli_app<'a, 'b>() -> App<'a, 'b> {
     App::new("beacon_node")
@@ -188,6 +189,20 @@ pub fn cli_app<'a, 'b>() -> App<'a, 'b> {
                 .takes_value(true)
                 .default_value("http://localhost:8545")
         )
+        .arg(
+            Arg::with_name("slots-per-restore-point")
+                .long("slots-per-restore-point")
+                .value_name("SLOT_COUNT")
+                .help("Specifies how often a freezer DB restore point should be stored. \
+                       DO NOT CHANGE AFTER INITIALIZATION.")
+                .takes_value(true)
+                .default_value(
+                    Box::leak(
+                        format!("{}", StoreConfig::default().slots_per_restore_point)
+                            .into_boxed_str()
+                    )
+                )
+        )
         /*
          * The "testnet" sub-command.
          *