[SPARK-31900][SPARK-SUBMIT] Client memory passed unvalidated to the JVM Xmx

[gerashegalov]

An example of a command that fail in the client mode but would not in the cluster mode or if the same value were used for spark.executor.memory

> $SPARK_HOME/bin/spark-shell --conf spark.driver.memory=" 4g "
Invalid maximum heap size: -Xmx 4g
Error: Could not create the Java Virtual Machine.
Error: A fatal exception has occurred. Program will exit.

also note that the same message is produced by the JVM makes it hard to identify which of the configs SPARK_DAEMON_MEMORY, SPARK_DRIVER_MEMORY, etc was responsible

Why are the changes needed?

easier ops/diagnostics, more uniform behavior

Does this PR introduce any user-facing change?

fix-only change

How was this patch tested?

• unit tests
• manual testing

[gatorsmile]

ok to test

[SparkQA]

Test build #123546 has finished for PR 28720 at commit 87e1d67.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[dongjoon-hyun]

Retest this please.

[SparkQA]

Test build #124240 has finished for PR 28720 at commit 87e1d67.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[github-actions]

We're closing this PR because it hasn't been updated in a while. This isn't a judgement on the merit of the PR in any way. It's just a way of keeping the PR queue manageable.
If you'd like to revive this PR, please reopen it and ask a committer to remove the Stale tag!

[srowen]

This is a whole lot of extra code to catch one particular type of typo, which even produces an error that narrows it down pretty far. I'm not sure it's worthwhile.

[gerashegalov]

thanks for looking at the PR @srowen. Since there are multiple culprits potentially responsible for the issue I thought a patch was warrant. And I like to see consistent behavior between similar settings. However, I also understand the tradeoff of accepting more code for future maintenance.