Better behavior for new messages alert

[tmaybe]

If there are only messages from one client, clicking the alert will take you directly to that client's message stream. If there are messages from more than one client, you're taken to the client list.

Also, the number of unread messages are now part of the text of the alert.

Closes #384

[codecov-io]

Codecov Report

Merging #409 into master will increase coverage by 3.18%. The diff coverage is 88.42%.

@@            Coverage Diff             @@
##           master     #409      +/-   ##
==========================================
+ Coverage   63.78%   66.96%   +3.18%     
==========================================
  Files         112      111       -1     
  Lines        4299     4262      -37     
  Branches      473      460      -13     
==========================================
+ Hits         2742     2854     +112     
+ Misses       1557     1408     -149

Impacted Files	Coverage Δ
deploy/chef/cookbook/files/default/credentials.js	76.92% <ø> (-1.65%)	:arrow_down:
app/controllers/alerts.js	62% <100%> (+8%)	:arrow_up:
seeds/seeds.js	97.7% <100%> (+0.59%)	:arrow_up:
app/controllers/access.js	18.91% <50%> (ø)	:arrow_up:
app/models/messages.js	60.57% <58.33%> (+15.48%)	:arrow_up:
app/lib/sms-status-check.js
app/controllers/voice.js	94.33% <0%> (+2.51%)	:arrow_up:
app/lib/models.js	92.39% <0%> (+4.34%)	:arrow_up:
app/models/conversations.js	75.4% <0%> (+4.91%)	:arrow_up:
... and 3 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 4cfc3bf...7fbe855. Read the comment docs.

[glassresistor]

I think in general the changes look good and am only hesitant if we are slowing down the request and/or adding significant queries,. @tdooner point about hitting the endpoint every 1s is salient. I think we can definitely tune that down to ~5sec or slower for a quick solution.

If we have more time we can work a way to keep count of unread messages in a way thats fast to read regularly.

In terms of testing if we can't test the frontend js then I agree we should have it just take a count for now.

[tmaybe]

Thank you! I think that, while the polling may be inefficient, the changes don't make it more inefficient. Merging now.

[tdooner]

Agreed, and this is still a hypothetical performance problem. The CPU usage on the web servers is negligible, last time I checked.

When we rewrite it in Rails, we can give ActionCable a shot 😀

[tmaybe]

ooh, I was wondering if there was a socket.io for rails :)