Make aws_instance_invalid_ami rule faster

[wata727]

Currently, aws_instance_invalid_ami rule fetches all images first. However, these are very large and take a lot of time to complete the request. We just had to provide the --fast option just to disable this rule.

In this pull request, instead of fetching all images first, it will request each time as needed, and cache the list of fetched AMIs. This can reduce the time spent on requests and significantly reduce memory usage.

The number of requests simply increases, but in many cases, it is faster than fetching all images. Here is an example of time output:

Before

tflint --deep  46.28s user 10.51s system 63% cpu 1:29.18 total

After

tflint --deep  0.55s user 0.10s system 8% cpu 7.817 total

NOTE: The --fast option will be no longer used, and will be removed in v0.9.