[SPARK-40948][SQL][FOLLOWUP] Restore PATH_NOT_FOUND

[itholic]

What changes were proposed in this pull request?

The original PR to introduce the error class PATH_NOT_FOUND was reverted since it breaks the tests in different test env.

This PR proposes to restore it back.

Why are the changes needed?

Restoring the reverted changes with proper fix.

Does this PR introduce any user-facing change?

No

How was this patch tested?

The existing CI should pass.

[LuciferYang]

Is the sparkr UTs failure is related to this one?

https://github.com/itholic/spark/actions/runs/3425639144/jobs/5708796073

══ Failed ══════════════════════════════════════════════════════════════════════
── 1. Failure ('test_sparkSQL.R:3993'): Call DataFrameWriter.load() API in Java 
`read.df("arbitrary_path")` threw an error with unexpected message.
Expected match: "Error in load : analysis error - Path does not exist"
Actual message: "Error in load : analysis error - [PATH_NOT_FOUND] Path does not exist: file:/__w/spark/spark/arbitrary_path."
Backtrace:
  1. testthat::expect_error(read.df("arbitrary_path"), "Error in load : analysis error - Path does not exist")
       at test_sparkSQL.R:3993:2
  6. SparkR::read.df("arbitrary_path")
  7. SparkR:::handledCallJMethod(read, "load")
  8. base::tryCatch(...)
  9. base (local) tryCatchList(expr, classes, parentenv, handlers)
 10. base (local) tryCatchOne(expr, names, parentenv, handlers[[1L]])
 11. value[[3L]](cond)
 12. SparkR:::captureJVMException(e, method)

[LuciferYang]

Looks like we need to refactor this case

spark/R/pkg/tests/fulltests/test_sparkSQL.R

Lines 3986 to 3998 in c5d2760

	test_that("Call DataFrameWriter.load() API in Java without path and check argument types", {
	# This tests if the exception is thrown from JVM not from SparkR side.
	# It makes sure that we can omit path argument in read.df API and then it calls
	# DataFrameWriter.load() without path.
	expect_error(read.df(source = "json"),
	paste("Error in load : analysis error - Unable to infer schema for JSON.",
	"It must be specified manually"))
	expect_error(read.df("arbitrary_path"), "Error in load : analysis error - Path does not exist")
	expect_error(read.json("arbitrary_path"), "Error in json : analysis error - Path does not exist")
	expect_error(read.text("arbitrary_path"), "Error in text : analysis error - Path does not exist")
	expect_error(read.orc("arbitrary_path"), "Error in orc : analysis error - Path does not exist")
	expect_error(read.parquet("arbitrary_path"),
	"Error in parquet : analysis error - Path does not exist")

[MaxGekk]

@HyukjinKwon @srielau @cloud-fan Are you ok with such changes?

[itholic]

FYI: the arbitrary_path here is different per test environments, so I use regexp for the path string.

[cloud-fan]

LGTM

[HyukjinKwon]

thanks man

[MaxGekk]

+1, LGTM. Merging to master.
Thank you, @itholic and @HyukjinKwon @cloud-fan @LuciferYang for review.