feat: Add APIs for updating Protocol and Metadata in Transaction

[lonless9]

What changes are proposed in this pull request?

closes #1055

This PR extends the Transaction API to support updating the table's Protocol and Metadata.

Added update_protocol(Protocol) and update_metadata(Metadata) methods to the Transaction struct.

How was this change tested?

unit tests

[codecov]

Codecov Report

❌ Patch coverage is 83.22981% with 27 lines in your changes missing coverage. Please review.
✅ Project coverage is 84.53%. Comparing base (e339ebf) to head (b8fb161).

Files with missing lines	Patch %	Lines
kernel/src/transaction/mod.rs	81.11%	10 Missing and 17 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1523      +/-   ##
==========================================
- Coverage   84.53%   84.53%   -0.01%     
==========================================
  Files         123      123              
  Lines       33279    33439     +160     
  Branches    33279    33439     +160     
==========================================
+ Hits        28133    28268     +135     
- Misses       3763     3771       +8     
- Partials     1383     1400      +17     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[zachschuermann]

hi @lonless9 thank you for the contribution! unfortunately there is a bit of nuance to these changes. protocol/metadata updates will require some more 'orchestration' than maybe is first apparent. This is since there is a dependency between data writes and metadata updates. By example: imagine a user starts a transaction, writes a bunch of data by getting the write_context, and then updates the table schema. The new table schema could be totally incompatible with the writes that just occurred - thereby leading to an incorrect write. In the short term, we could enforce all metadata updates occurring prior to handing out any WriteContext but I think @sanujbasu is in the middle of some Transaction API re-works to allow for better enforcement of these invariants more broadly.

I'll maybe let @sanujbasu chime in with the best path forward here given you both have some changes in flight?

[lonless9]

Thanks for the review! @zachschuermann

I previously assumed that in practical usage scenarios, obtaining the write context and determining metadata updates would naturally happen within a consistent context. Now I see your point: because Kernel decouples these steps, it indeed creates a risk where incorrect user code could happen.

I agree that enforcing these invariants at the API level is a good short-term approach. Per your suggestion, let's see what @sanujbasu thinks!