Add more test cases for submitBlocksToDA in manager_test.go

[Manav-Aggarwal]

This needs an update to testing. 

Test cases I can think of that we should understand how they execute. 
Assuming starting with Blocks a b c
1. happy case, all are submitted on first round
2. a and b are submitted first because including c triggers size limit. C is then submitted on second round.
3. C is too big on it's own, never submitted but a and b are
4. C has a marshal issue. A and b should still be submitted ( I think this case currently isn't possible with the code based on `SubmitBlocks`)
5. A is too big on its own 
6. A has a marshal issue

for cases 5 and 6, what should be the desired outcome? Do we never submit any blocks if the current next block is bad? Based on `SubmitBlocks` that appears to be the case.

_Originally posted by @MSevey in https://github.com/rollkit/rollkit/pull/1471#pullrequestreview-1832928038_

Note that test cases 1, 2, 3 here are added and test cases 4, 5, 6 need to be added.

[arhamj]

Can we catch serialisation errors much earlier at like the addPendingBlock stage. In the current implementation, this would come at the cost of serialising the same block an extra time but, we could avoid that by caching the result from the early serialisation step. This change could potentially help keep the pending block struct smaller.

I'll proceed with writing the tests for cases 4, 5 and 6 for now.

[arhamj]

Another option would be to separate the aggregation of blocks from the submission to DA and have the method which does the aggregation (applying size limits and serialization) in the block manager. Eviction on error could happen at this stage easily. (Probably log and evict for now?)