CDF-20156: fixup fatjar pom file (#858)

Currently
https://repo1.maven.org/maven2/com/cognite/spark/datasource/cdf-spark-datasource-fatjar_2.13/3.2.1044/cdf-spark-datasource-fatjar_2.13-3.2.1044.pom
includes
```xml
<dependencies>
  <dependency>
    <groupId>org.scala-lang</groupId>
    <artifactId>scala-library</artifactId>
    <version>2.13.8</version>
  </dependency>
  <dependency>
    <groupId>com.cognite.spark.datasource</groupId>
    <artifactId>cdf-spark-datasource_2.13</artifactId>
    <version>3.2.1044</version>
  </dependency>
  <dependency>
    <groupId>io.scalaland</groupId>
    <artifactId>chimney_2.13</artifactId>
    <version>0.5.3</version>
  </dependency>
</dependencies>
```
Which isn't expected, fatjar should have (almost) no dependencies.
The jar itself though contains all the classes and applies shading
as expected.
It is then possible that pulling fatjar from maven central would pull duplicated classes
and if non-fatjar unshaded versions are picked up or a mix with fatjar ones, the
datasource would runtime-crash when used for example in Spark in DataBricks.

Let's remove chimney from commonSettings and move to per-target dependencies.
And for main library, looks like the simplest way to both
have it as part of fatJar and not have it in pom is to patch generated pom.

As for scala-library keeping it as-is, older fatjar setup also had it,
so not investigating for now.

Verified pom contents by running `fatJarShaded/publishLocal` and inspecting pom
from filesystem.
And used `fatJarShaded/assembly` to generate fatjar and inspect with `jar tf`
to check that is has our classes and applies shading rules.
Did not yet verify in databricks, but even if this PR isn't enough to fix
running there it is still a step in the right directiion.

[CDF-20156]

build.sbt

@@ -1,6 +1,8 @@
 import com.typesafe.sbt.packager.docker.Cmd
 import sbtassembly.AssemblyPlugin.autoImport._
 import sbtassembly.MergeStrategy
+import scala.xml.{Node => XmlNode, NodeSeq => XmlNodeSeq, _}
+import scala.xml.transform.{RewriteRule, RuleTransformer}
 
 val scala212 = "2.12.15"
 val scala213 = "2.13.8"
@@ -26,7 +28,7 @@ lazy val commonSettings = Seq(
   organization := "com.cognite.spark.datasource",
   organizationName := "Cognite",
   organizationHomepage := Some(url("https://cognite.com")),
-  version := "3.3." + patchVersion,
+  version := "3.4." + patchVersion,
   isSnapshot := patchVersion.endsWith("-SNAPSHOT"),
   crossScalaVersions := supportedScalaVersions,
   semanticdbEnabled := true,
@@ -35,7 +37,6 @@ lazy val commonSettings = Seq(
   description := "Spark data source for the Cognite Data Platform.",
   licenses := List("Apache 2" -> new URL("http://www.apache.org/licenses/LICENSE-2.0.txt")),
   homepage := Some(url("https://github.com/cognitedata/cdp-spark-datasource")),
-  libraryDependencies ++= Seq("io.scalaland" %% "chimney" % "0.5.3"),
   scalacOptions ++= Seq("-Xlint:unused", "-language:higherKinds", "-deprecation", "-feature"),
   resolvers ++= Seq(
     Resolver.sonatypeRepo("releases")
@@ -91,6 +92,7 @@ lazy val structType = (project in file("struct_type"))
     name := "cdf-spark-datasource-struct-type",
     crossScalaVersions := supportedScalaVersions,
     libraryDependencies ++= Seq(
+      "io.scalaland" %% "chimney" % "0.6.1",
       "org.typelevel" %% "cats-core" % "2.9.0",
       "org.apache.spark" %% "spark-sql" % sparkVersion % Provided,
     ),
@@ -186,6 +188,24 @@ lazy val fatJarShaded = project
       )
     },
     assembly / assemblyOption := (assembly / assemblyOption).value.withIncludeScala(false),
+    pomPostProcess := { (node: XmlNode) =>
+      new RuleTransformer(new RewriteRule {
+        override def transform(node: XmlNode): XmlNodeSeq = node match {
+          case e: Elem if e.label == "dependency"
+                          && e.child.filter(_.label == "groupId").flatMap(_.text).mkString == "com.cognite.spark.datasource"
+                          && e.child.filter(_.label == "artifactId").flatMap(_.text).mkString.startsWith("cdf-spark-datasource") =>
+            // Omit library artifact from pom's dependencies.
+            // All sbt-assembly settings are kept here and we can't depend on
+            //   Compile / packageBin := (library / assembly).value
+            // as it would try to run sbt-assembly on library too and it doesn't have
+            // all the configuration.
+            // Otoh library itself should remain pure non-fatjar library and know nothing about sbt-assembly.
+            // There could be other ways to achieve the same effect, so far this one is found and it is simple enough.
+            Seq()
+          case _ => node
+        }
+      }).transform(node).head
+    }
   )
 
 addCompilerPlugin("com.olegpy" %% "better-monadic-for" % "0.3.1")