http 500 when request is too large

[Dieterbe]

since grafana/metrictank@a63233c
this can happen:

T 10.240.78.52:38934 -> 10.240.0.6:18763 [AP]
GET /get?to=1456360072&from=1424824072&target=7.78be844af2ee3b995158deabd10f5b52&target=7.dd44b380fac317e139c7c7a7a0e0173f&target=7.8e38e7d1c4f68fe02b269141003b5689&target=7.18c480f0d23fe882c
a1c9a8a7894b3ee&target=7.5fdf53acb83850604c3473dad1f27ffe&target=7.5d4a9ebdde475e14e266b95f2e35be5f&target=7.ec9ac5ee71515ad0854f77254f0d2653&target=7.795b43d157f110ca80c16bbc8da16734&target=
7.939766ffdab733bf5fd7e16641d7dfc7&target=7.61226aea7ae4e02a4f7dc6ac2f21886b&target=7.5b8c263e6c821168f610d055d0209198&target=7.d2376d1b65ac56d0d3539fa0f9855efd&target=7.d14a29408a9f1cd0d556b
4b1b96411d1&target=7.b7a7a4ae2204f375937b5771101a4cb2&target=7.95eb512754370df5f6ec20ed5a49a3bc&target=7.e94eb05688a6d6e4956ff92a5b89591e&target=7.037539605c8b6b031fc6c4c2afe5ac07&target=7.791b300744d4b73c3ec97c4a16cfe74a&target=7.ef8682e9bd04640ce949eca83b47c457&target=7.5cea4e37327311065c9b44526d9faedf&maxDataPoints=480 HTTP/1.1.
Host: 10.240.78.52:18763.
Connection: keep-alive.
Accept-Encoding: gzip, deflate.
Accept: */*.
User-Agent: graphite_raintank.
.

...
##
T 10.240.0.6:18763 -> 10.240.78.52:38934 [AP]
HTTP/1.1 400 Bad Request.
Content-Type: text/plain; charset=utf-8.
X-Content-Type-Options: nosniff.
Date: Thu, 25 Feb 2016 00:27:52 GMT.
Content-Length: 24.
.
too much data requested

I think:

• adjust time selection options grafana#569
• I'll also update MT to be more permissive
• graphite-raintank should pass the error through more elegantly, so it is shown in grafana.
  because people will be able to get in this situation fairly easily.
  currently, graphite will http 500 and show ValueError: No JSON object could be decoded and a stack trace, and in grafana it looks like an error with the ugly stack trace.
  note also that Content-Type is set to text/plain, and not application/json. if MT returns http 400 with an error, graphite-raintank should just return 400 with the same error.