ci: Updated security agent workflow to use large runners when available

[bizob2828]

Description

We've had flappy nightly runs with versioned tests with security agent and memcached tests.

 # Subtest: captures datastore instance attributes with multiple hosts
        # Subtest: separate gets
            not ok 1 - timeout!
              ---
              expired: memcached instrumentation
              test: separate gets
              ...
            
            1..1
            # failed 1 test
        not ok 1 - separate gets # time=36.22ms
        
        # Subtest: multi-get
            not ok 1 - Maximum call stack size exceeded
              ---
              stack: |
                Client.stubbedCommand (memcached.tap.js:654:9)
                Client.stubbedCommand (memcached.tap.js:654:21)
                Client.stubbedCommand (memcached.tap.js:654:21)
                Client.stubbedCommand (memcached.tap.js:654:21)
                Client.stubbedCommand (memcached.tap.js:654:21)
                Client.stubbedCommand (memcached.tap.js:654:21)
                Client.stubbedCommand (memcached.tap.js:654:21)
                Client.stubbedCommand (memcached.tap.js:654:21)
                Client.stubbedCommand (memcached.tap.js:654:21)
                Client.stubbedCommand (memcached.tap.js:654:21)
              at:
                line: 654
                column: 9
                file: memcached.tap.js
                function: Client.stubbedCommand
              type: RangeError
              tapCaught: testFunctionThrow
              test: multi-get
              source: |2
                        /* eslint-enable no-unused-vars */
                        origCommand.call(this, queryCompiler, realServer)
                --------^

I suspect it's the limitations of running with the default runner. This PR uses the larger runner.