Construct dependency fingerprint from canonicalized hash tree

[greglook]

During some debugging, I found that the current dependency fingerprinting code is hashing the dependency structure basically as-is, leaving it brittle against changes in things like dependency ordering and map iteration order. Instead, we should canonicalize each dependency coordinate and roll things up into a tree of hashes.

A little debug printing shows that a dependency calculation for the current code is hashing lots of raw data structures:

Hashed input:                                                                                                                                                                                                                                        
v2/dependencies                                                                                                                                                                                                                                      
:lein-monolith.task.fingerprint/default {:dependencies ([org.clojure/clojure "1.11.1"] [amperity/aws "MONOLITH-SNAPSHOT" :exclusions ([com.amazonaws/*])] ,,, [mvxcvi/puget nil] [mvxcvi/whidbey nil])}
:monolith/leaky {:managed-dependencies [[aleph "0.4.6" :exclusions [byte-streams]] [amperity/envoy "1.0.1"] ,,, [yada/lean "1.2.16" :exclusions [byte-streams]]]}
:provided       {:dependencies ([org.apache.spark/spark-core_2.12 "3.1.2" :exclusions ([org.apache.hadoop/hadoop-client])] [org.apache.spark/spark-sql_2.12 "3.1.2"] [com.amazonaws/aws-java-sdk-bundle "1.11.1026"] ,,,)}
    => 595wMrKcf8UPCOpWyl032pohkh0

After this change, the same debugging shows the tree rollup:

...

Hashed input:
v2/dep-coordinate
:exclusions	([com.amazonaws/*])
:version	"MONOLITH-SNAPSHOT"
    => IvYsPCjF3X2QDGuKqFMvJOM6tLs

Hashed input:
v2/dep-coordinate
:version	"MONOLITH-SNAPSHOT"
    => WnEIDZX_gfYyKNWTg1EwmZTXPhg

Hashed input:
v2/dep-coordinate
:version	nil
    => 6h06eUyN4JXUfcpzZFS-zXwmZsE

...

Hashed input:
v2/dep-coordinate
:version	"0.6.0"
    => AekLXJrkY2zBH8TGKcsGTd1IYRE

Hashed input:
v2/dep-coordinate
:version	"3.1.2"
    => t16552m11qHRfTSpB5sVKnGoTH0

Hashed input:
v2/dep-coordinate
:version	"1.11.1"
    => x8Jetn5-OcsZWy59Spfvr2I7n7Y

Hashed input:
v2/profile-dependencies
amperity/aws	IvYsPCjF3X2QDGuKqFMvJOM6tLs
amperity/domain-core	WnEIDZX_gfYyKNWTg1EwmZTXPhg
amperity/envoy	6h06eUyN4JXUfcpzZFS-zXwmZsE
amperity/hadoop	tj1lCqROWc2_Z_JXwOpUAqTCocg
amperity/spark-base	qoKQMZNzrGA1wNeC81DluNOpBSA
amperity/spark-env	WnEIDZX_gfYyKNWTg1EwmZTXPhg
amperity/storage-coordinates	WnEIDZX_gfYyKNWTg1EwmZTXPhg
amperity/storage-spark	WnEIDZX_gfYyKNWTg1EwmZTXPhg
cider/cider-nrepl	ixE63Qpr_sc5kjv97PQNRW6vDaw
com.stuartsierra/component	6h06eUyN4JXUfcpzZFS-zXwmZsE
mvxcvi/puget	6h06eUyN4JXUfcpzZFS-zXwmZsE
mvxcvi/whidbey	6h06eUyN4JXUfcpzZFS-zXwmZsE
nrepl/nrepl	AekLXJrkY2zBH8TGKcsGTd1IYRE
org.apache.spark/spark-hadoop-cloud_2.12	t16552m11qHRfTSpB5sVKnGoTH0
org.clojure/clojure	x8Jetn5-OcsZWy59Spfvr2I7n7Y
    => tBPKOsU5ejRVwjDlYD8MIUlq35g

Hashed input:
v3/dependencies
:lein-monolith.task.fingerprint/default	:dependencies	tBPKOsU5ejRVwjDlYD8MIUlq35g
:monolith/leaky	:managed-dependencies	XDQPF6EzKm2Vg-NKeC9AXDcGg9c
:provided	:dependencies	wOCcXLakP6MwREX_PbOUNnT7iaI
    => fQZgHkrH2g-Ap-GQtv_pU0Sbbxo

I tested and reordering dependencies within a project no longer changes the dependency fingerprint for that project.