database: error instead of dying on monitor failure

[g-k]

Code changes:

• change DB monitor to log errors instead of dying (only the csigpki signer uses the DB and it sets up and loads its config at init, so it's OK to briefly lose DB connectivity and we shouldn't crash the app)
• factor out DB init to addDB in main
• factor out CheckConnection func
• add connect/monitor tests
• add MonitorPollInterval to DB config
• distinguish the heartbeat handler from LB heartbeat and teach it to check HSM connectivity (it finds the first signer config with a non-PEM private key, caches it, and uses it to check HSM connectivity returning 502 bad gateway when the connection fails to remove it from the health LBs cc @milescrabill @Micheletto re: this behavior)

Unit tests just check DB connected and HSM disconnected, but I functional tested that running app-hsm and hitting http://localhost:8001/__heartbeat__ returns

{"hsmAccessible": true}

r? @jvehent

TODO assuming this lands:

• monitor logs for HSM and DB disconnects
• add monitorpollinterval to autograph configs

[coveralls]

Pull Request Test Coverage Report for Build 2981

• 64 of 128 (50.0%) changed or added relevant lines in 4 files are covered.
• 1 unchanged line in 1 file lost coverage.
• Overall coverage decreased (-0.1%) to 70.86%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
database/connect.go	17	21	80.95%
handlers.go	34	43	79.07%
signer/signer.go	13	28	46.43%
main.go	0	36	0.0%

Files with Coverage Reduction	New Missed Lines	%
main.go	1	45.08%

Totals
Change from base Build 2955:	-0.1%
Covered Lines:	3171
Relevant Lines:	4475

---

💛 - Coveralls

[jvehent]

r+, with some minor comments