Add resize_price_account support; create .dockerignore

[drozdziak1]

Motivation

This is the main changeset responsible for price account resizing in program-admin. Account resizing is realized as a new CLI subcommand (resize-price-accounts-v2). The new command performs account discovery similar to sync, finds accounts with insufficient data size and expands them using the relevant on-chain instruction. The reason for a separate subcommand (and not expanding sync) is that resizing will likely be performed only once per oracle deployment, as with current program-admin code, new price accounts are already allocated to use expanded PriceAccountV2 size.

In order to perform tests in integration, a way to allocate v1-sized price accounts was necessary. To support that, the sync subcommand got an option to allocate new price accounts using v1 size (--allocate-price-v2, uses v1 size when set to false). This way, when resize_price_account is called on chain, a true resize is performed. The resize would be skipped otherwise, as the on-chain instruction exits early if there's enough space already.

Additionally, a .dockerignore file is created from .gitignore to prevent Python runtime artifacts and locally-stored secrets from being copied into Docker storage.

Summary of changes

• cli.py - Add new subcommand, add choice between v1 and v2 account size for sync price account allocation
• __init__.py - Add logic for finding, filtering and expanding v1 price accounts;
• instructions.py - Implement resize_price_account on-chain instruction
• util.py - Differentiate between V1 and V2 account sizes in constants
• parsing.py - Make parse_price aware of the extra price components for v2 prices, ensure that trailing PriceCumulative data is not misinterpreted as a price component
• pyproject.toml - Bump package version to v0.1.2
• .dockerignore - copied from .gitignore

Testing

• The new subcommand works well within the integration environment, correctly identifying v1 accounts and v2 ones. The resize transactions succeed and do not break other services in the Tilt environment (agent continues to publish).
• Existing unit tests continue to pass

Review highlights

• parsing.py - I decided against parsing the trailing data after expanded price component array, because the data is not relevant to administrative tasks on price accounts. It is indicated in the code comments.

[jayantk]

seems fine to me, though I can't say whether or not the txs are constructed exactly right. This could definitely use some tests, but maybe those go in a different repo? I've sort of lost the thread on the testing story here.

Anyway, I'm going to approve this, but please do add tests as a follow up.

[thmzlt]

I guess the new code is covered because the flag value defaults to true.

[drozdziak1]

Changes since last review

• parsing.py - Opted for using the existing num_components field for price component parsing, which I previously missed. This greatly simplifies the change to parsing.py, making extensive testing less necessary.
• test_sync.py - changed all instances of sync_from_files to use allocate_price_v2=False - This gives opportunity to parsing.py code to encounter shorter v1 buffers, ensuring that parsing never goes out of bounds
• .dockerignore, .gitignore - Exclude Python package builds from Docker and Git, located in dist/