You can put your zones.csv into a resources subfolder indeed, this way it will stay with your measure (it seems that you can only have measure.rb and measure.xml at the root of your measure folder).

In your zones.csv:

ZoneName
Zone 1
Zone 2
Zone 3

then in your measure:

require 'csv'
# File.dirname(__FILE__) points to the location of the current file, so your `measure.rb`
zones = CSV.table(File.dirname(__FILE__) + '/resources/zones.csv')
#

At this point if you check the headers:

[1] (main)> zones.headers
 => [:zonename]

Then you can do a bunch of stuff, like you can loop on each and grab the thermal zone associated with it

zones.each do |row|
    zonename = row[:zonename]
    zone = model.getThermalZoneByName(zonename)
    if zone.empty?
        puts "couldn't find zone named #{zonename}"
    else
        zone = zone.get
        # Do something else
    end
end

Or if it's easier you can always load all the zone name in a more familiar array object:

# Create empty array to store the zone names
 zone_names = []
 zones.each{|row| zone_names << row[:zonename]}

Adding to @Julien Marrec's answer:

There's not currently a path argument available to OS measures, but you can approximate that function with a simple string argument:

#make an argument for location of .csv file
csv_file_path = OpenStudio::Ruleset::OSArgument::makeStringArgument("csv_file_path",true)
csv_file_path.setDefaultValue("C:\Users\Eric\Desktop\zones.csv")
args << csv_file_path

Then in the run section, apply that string to a variable that you pass into the CSV method of your choice.

Just to through in more options. If your list isn't going to be too long, you can have a string argument that expects comma separated values in it, instead of using an external file.

Then in the measure you parse it apart using code like this.

array_from_arg = arg.split(',')

Here is another option that is possible but not really recommended - The arguments section of the measure can loop through zones and make a bool argument for each zone. Then you can check the zones you want included. This isn't ideal because having a variable/unknown number of arguments for a measure can cause issues with large scale analysis, and if you have a really big model this may blow up. If intended to used only with apply measures now, then this could be ok.