Upgrade collector to use concurrent futures and parallelize over loading conduit nodes

• Add type annotations to function calls
• Add a 3rd party linter to makefile to check for them (eg mypy)

The current serialize command can become verbose when converting a large serialized JSON string with nested subtrees into a JSON file. I propose adding an additional syntax to reduce the verbiage. Or even introduce a new command called deserialize that will convert a serialized tree into a JSON file.

Proposal:
  A new syntax added to serialize.
    or
  A new command called deserialize. I'm more inclined to this option since the name accurately describes the proposed behavior.

We will use the following JSON to demonstrate the differences;

{
  "people" : {
    "ryan" : {
      "age" : 23,
      "hobbies" : {
        "soccer" : "good",
        "running" : "bad"
      },
      "status" : "insomina"
    },
    "john" : {
      "age" : 45,
      "hobbies" : {
        "soccer" : "bad",
        "running" : "good"
      },
      "status": "sad"
    }
  }
}

Old Syntax:

spellbook serialize --output myfile.json --vars \
people/ryan/age=23 \
people/ryan/hobbies/soccer=good \
people/ryan/hobbies/running=bad \
people/ryan/status=insomania \
people/john/age=45 \
people/john/hobbies/soccer=bad \
people/john/hobbies/running=good \
people/john/status=sad\

New Syntax:

spellbook deserialize --output myfile.json \
people>\
ryan>age=23&hobbies>soccer=good&running=bad<status=insomania<\
john>age=45&hobbies>soccer=bad&running=good&status=sad\

Breaking Down the Syntax:

• ">" tells the program that the next value to read is a child node.
• "<" tells the program to return to the parent node.
• "&" tells the program that the next value to read is a sibling node.

Things to Note:

• After specifying "hobbies" and its child nodes, a "<" symbol is appended. The first instinct is to append "&," but a "<" symbol is needed to tell the program to return to the parent node of "hobbies." This will correctly place "status" as a sibling of "age" and "hobbies."
• The optional variable "--vars" is not needed since the serialized tree is one continuous string with no spaces.

🐛 Bug Report

Describe the bug
When running make-samples with rounding in a merlin workflow, the following error occurs:

[2023-03-21 11:36:34: INFO] Generating samples...
[2023-03-21 11:36:37: INFO] Generating samples complete!
[2023-03-21 11:36:37: INFO] Loading samples from 'samples.npy'...
[2023-03-21 11:36:37: ERROR] Object arrays cannot be loaded when allow_pickle=False

To Reproduce
Steps to reproduce the behavior:

1. Create a merlin workflow and use the make-samples command with roudning in the "samples" step.
2. Run merlin workflow in dry mode.

Expected behavior
Merlin should be able to load the samples.

Additional context
When not using the -round option, the workflow is able to load the samples.

I'm using the following packages:

merlin-spellbook==0.8.1
merlin==1.8.3 (not using 1.9.1 b/c it is unable to correctly resubmit itself when doing an iterative workflow)

Command
spellbook make-samples -dims 6 -n 8 -sample_type lhs -outfile=samples.npy -scale "[[-0.5,0.5],[-0.5,0.5],[430,452],[60,70],[50,60],[0,1]]" -seed 10 -round "[False,False,False,False,False,round]"