read more distributions

[rshewitt]

related to 5767

• no longer limits resources/distributions based on available formats. this grabs all available ones within the expected context.
• also fixes a bug which could produce duplicate resources (i haven't seen this happen but this is a good fix nevertheless)
• there's a possibility for resources to be excluded if there's no distributor (it's an optional element in the source) see below.

<gmd:MD_Distribution>
  <!-- no distributor element --> 
  <gmd:transferOptions> 
    <!-- ... --> 
  </gmd:transferOptions>
</gmd:MD_Distribution>

in this ^ circumstance there's a resource/distribution but because there's no distributor and the internal md object doesn't support distributor-less distributions this would return 0 resources. in my entire time of working on mdtranslator i've never had to update the internal md object so i want to avoid that as much as possible. if there's lots of documents in this scenario then i suppose we could create a dummy distributor to put the resources into but I would encourage the data provider to include one instead.

accessURL & downloadURL

the main purpose of this PR is to bring in more distributions but while i'm also bringing back mimetype calculation in harvester I decided to correct how we process accessURL and downloadURL. the logic for determining those fields isn't accurate. for example, an arcgis map server distribution/resource with a link https://tigerweb.geo.census.gov/arcgis/rest/services/TIGERweb/tigerWMS_Current/MapServer would be considered a downloadURL but on ckan catalog it's a web page. this noaa dataset on beta has 14 resources as downloadable files. that's incorrect. they should be web pages to visit. this PR fixes that.