Dark order and alert persistence

[algorandpa]

We should be able to submit a dark order and/or alert that does not clear, stop and start pikerd, ensure order is both serialized to a local db (likely file system) on pikerd shutdown and reloaded and re-registered in emsd state and the dark clearing loop on restart.

---

Design and impl ToDo:

• implement functionality to persist dark-orders by serializing Order into a toml style format much like how we currently store positions and trade ledgers.
• add an api to both read and write Order structs to the filesystem as well as startup/teardown code inside the emsd actor to sync the clearing engine state with any offline stored entries

---

Testing ToDo:

• tests to verify submitted-non-cleared darks and alerts are written to file after ems shutdown
• tests to verify dark order / alert state on reboot of emsd all via client APIs

[goodboy]

@algorandpa probably the stuff proposed in terms of better TOML infra / libs in #345 is worth linking here.

[goodboy]

Oh, also the same idea / storage format should likely be applied for reloading "live paper" limits that were submitted but should get reloaded after the paper engine has gone through a full power cycle 😂