[MINOR] build: Move dashboard module into profile

[EnricoMi]

What changes were proposed in this pull request?

The dashboard module should be moved into a profile and and not build by default.

It can be activated via -Pdashbboard.

Why are the changes needed?

Building and especially incrementally compiling during development always builds dashboard completely. All other modules finish in seconds, but dashboard takes minutes. This increases the iteration cycle from seconds to minutes.

Compiling Uniffle with dashboard:

./mvnw compile -Pdashboard
…
[INFO] ------------------------------------------------------------------------
[INFO] Reactor Summary for Apache Uniffle 0.9.0-SNAPSHOT:
[INFO] 
[INFO] Apache Uniffle ..................................... SUCCESS [  0.804 s]
[INFO] Apache Uniffle Protocols ........................... SUCCESS [  4.486 s]
[INFO] Apache Uniffle Common .............................. SUCCESS [  1.778 s]
[INFO] Apache Uniffle Internal Client ..................... SUCCESS [  0.244 s]
[INFO] Apache Uniffle Coordinator ......................... SUCCESS [  0.492 s]
[INFO] Apache Uniffle Storage ............................. SUCCESS [  1.433 s]
[INFO] Apache Uniffle Server .............................. SUCCESS [  1.161 s]
[INFO] Apache Uniffle Client .............................. SUCCESS [  0.381 s]
[INFO] Apache Uniffle CLI ................................. SUCCESS [  0.107 s]
[INFO] Apache Uniffle Integration Test (Common) ........... SUCCESS [  1.066 s]
[INFO] Apache Uniffle Dashboard ........................... SUCCESS [02:22 min]
[INFO] ------------------------------------------------------------------------
[INFO] BUILD SUCCESS
[INFO] ------------------------------------------------------------------------
[INFO] Total time:  02:35 min

Compiling Uniffle without dashboard:

./mvnw compile
…
[INFO] ------------------------------------------------------------------------
[INFO] Reactor Summary for Apache Uniffle 0.9.0-SNAPSHOT:
[INFO] 
[INFO] Apache Uniffle ..................................... SUCCESS [  0.820 s]
[INFO] Apache Uniffle Protocols ........................... SUCCESS [  4.560 s]
[INFO] Apache Uniffle Common .............................. SUCCESS [  1.825 s]
[INFO] Apache Uniffle Internal Client ..................... SUCCESS [  0.213 s]
[INFO] Apache Uniffle Coordinator ......................... SUCCESS [  0.496 s]
[INFO] Apache Uniffle Storage ............................. SUCCESS [  1.403 s]
[INFO] Apache Uniffle Server .............................. SUCCESS [  1.100 s]
[INFO] Apache Uniffle Client .............................. SUCCESS [  0.431 s]
[INFO] Apache Uniffle CLI ................................. SUCCESS [  0.106 s]
[INFO] Apache Uniffle Integration Test (Common) ........... SUCCESS [  1.052 s]
[INFO] ------------------------------------------------------------------------
[INFO] BUILD SUCCESS
[INFO] ------------------------------------------------------------------------
[INFO] Total time:  12.538 s
…


### Does this PR introduce _any_ user-facing change?
Only affects build time.

### How was this patch tested?
Manual testing.

[zuston]

LGTM. Nice improvement! @EnricoMi

[rickyma]

Currently, when running the build_distribution.sh script, it always compiles the dashboard, and there is no option to choose not to compile it. I think the build_distribution.sh script should support an option to not compile the dashboard.

WDYT? @EnricoMi

[EnricoMi]

I agree, it should also be removed from https://github.com/apache/incubator-uniffle/blob/d87dc900407d4faa9ba5f80cd19d08513e71082d/.github/workflows/parallel.yml#L87-L91
and added to the matrix profile instead.

Similarly, it should be removed from .github/workflows/sequential.yml and added as a an extra step.

[rickyma]

Could you continue working on this part when you have some free time? I would greatly appreciate it, this will be very helpful to me. @EnricoMi

[EnricoMi]

That just hit me as well, and my changes to build_distribution.s were very ignorant, so here is the fix #1525

[EnricoMi]

The CI is fixed in #1526. Please review.