feat(consumer): Emit batch_write_bytes metric from RowBinary writer

[onewland]

Summary

• Adds the insertions.batch_write_bytes counter to the RowBinary ClickHouse writer, matching what the JSONEachRow writer already emits
• This is Phase 0 of the byte-based batching rollout — we need to observe batch byte sizes on eap_items in s4s2/de (which use RowBinary) to pick a data-driven threshold for --max-batch-size-calculation=bytes

Refs EAP-460

Test plan

• cargo check passes
• Deploy to s4s2 and confirm insertions.batch_write_bytes appears in DataDog for RowBinary consumers
• Use DataDog arithmetic (bytes/msgs) to approximate per-batch byte size distributions across all environments

🤖 Generated with Claude Code

[linear-code]

EAP-460 Add max_batch_size_bytes to consumer

[sentry]

Bug: The insertions.batch_write_bytes metric for the RowBinary writer uses an in-memory size estimate (estimated_size()) instead of the actual serialized byte count, making it inconsistent with the JSONEachRow writer.
_{Severity: MEDIUM}

Suggested Fix

The RowBinary writer should measure the size of the data after it has been serialized into the RowBinary format, not before. This can be achieved by getting the length of the serialized byte buffer, which will ensure the metric reflects the actual number of bytes written and is consistent with the JSONEachRow writer's implementation.

Prompt for AI Agent

Review the code at the location below. A potential bug has been identified by an AI
agent.
Verify if this is a real issue. If it is, propose a fix; if not, explain why it's not
valid.

Location: rust_snuba/src/strategies/clickhouse/row_binary_writer.rs#L127-L133

Potential issue: The `insertions.batch_write_bytes` metric for the RowBinary writer is
calculated using `estimated_size()`, which measures the in-memory size of the data
structures. This is inconsistent with the JSONEachRow writer, which measures the actual
byte length of the serialized data. Because the RowBinary format uses variable-length
encoding, the in-memory size is not an accurate representation of the on-the-wire size.
This discrepancy will result in inaccurate and non-comparable metric data between the
two writers, which could lead to incorrect decisions about batching thresholds and
undermines the goal of standardizing this metric.

_{Did we get this right? 👍 / 👎 to inform future reviews.}