12728: Fix O(N) database updates during file upload on large datasets

[tuannx]

What this PR does / why we need it:
Fixes an $O(N)$ database write bottleneck in UpdateDatasetVersionCommand.java where adding a file or updating a dataset version unconditionally touched modificationTime on all existing files in the dataset.

In datasets with thousands of files (e.g. 14,000+ files), mutating every file's modification timestamp on each upload caused JPA dirty checking to execute an UPDATE DVOBJECT SET MODIFICATIONTIME = ... statement for each existing file on every flush ($\frac{N(N+1)}{2}$ cumulative updates). This caused per-file upload latency to degrade from 1s to over 30s.

The fix ensures UpdateDatasetVersionCommand only initializes timestamps on newly created files (dataFile.getCreateDate() == null), safely handles null timestamps on legacy files, or updates the target file when single-file variable metadata (fmVarMet) is explicitly passed.

Which issue(s) this PR closes:

• Closes Performance: O(N) database updates on every file upload degrade large dataset ingestion (1s -> 30s+) #12728

Special notes for your reviewer:
Includes:

1. src/main/java/edu/harvard/iq/dataverse/engine/command/impl/UpdateDatasetVersionCommand.java: Targeted fix to avoid dirtying existing files in JPA.
2. src/test/java/edu/harvard/iq/dataverse/engine/command/impl/UpdateDatasetVersionCommandTest.java: Acceptance tests asserting that:
   • Adding a file does NOT modify existing files' modificationTime (reproduces and fails on unpatched code, passes with fix).
   • Updating variable metadata only touches the target file.
   • Null modification times are safely initialized.
3. doc/release-notes/12728-perf-large-upload.md: Release notes following issue naming convention.

A standalone Docker reproduction benchmark environment and Python benchmark tool are provided in the companion repo:

• tuannx/dataverse-large-upload-repro

Suggestions on how to test this:
Run the JUnit 5 acceptance test:

mvn test -Dtest=UpdateDatasetVersionCommandTest

Or run the full command regression suite:

mvn test -Dtest=UpdateDatasetVersionCommandTest,CreateDatasetVersionCommandTest,AbstractDatasetCommandTest,UpdateDatasetLicenseCommandTest,UpdateDatasetFieldsCommandTest

Does this PR introduce a user interface change? If mockups are available, please link/include them here:
No.

Is there a release notes update needed for this change?:
Yes, added doc/release-notes/12728-perf-large-upload.md.

Additional documentation:

• Dataverse Community Google Groups: https://groups.google.com/g/dataverse-community/c/pzSjF1YPaJw/m/bQWX3W_kBwAJ?utm_medium=email&utm_source=footer
• Reproduction Environment: https://github.com/tuannx/dataverse-large-upload-repro

[pdurbin]

We discussed this at tech hours. Overall, we like the approach.

When it comes to how to test this, the following PR is related, and @landreev is actively using it for testing:

• Performance Testing Infrastructure Bits #12365

@tuannx also created this related PR:

• Add big batch sample dataset ingestion and benchmark tool dataverse-sample-data#48