feat(logs): add (u)int64 sentry_value_t type

[JoshuaMoelans]

Cherry-picked from #1297 (since we don't need the full alternative API for now). Merging into #1272 so we can send integers as int64.

Will initially be used mainly for the new Logs feature, but we should introduce (mostly test) this for other data we send too.

---

Introduces sentry_value_new_(u)int64()

---

#skip-changelog

[github-actions]

	Messages
📖	Do not forget to update Sentry-docs with your feature once the pull request gets approved.

Generated by 🚫 dangerJS against 84fb966

[JoshuaMoelans]

@sentry review

[seer-by-sentry]

On it! We are reviewing the PR and will provide feedback shortly.

[seer-by-sentry]

PR Description

This pull request introduces support for 64-bit integer types (signed and unsigned) within the Sentry native SDK's value system. This enhancement allows the SDK to accurately capture and transmit larger numerical values that exceed the range of the existing 32-bit integer type, improving data fidelity for applications dealing with such values.

Click to see more

Key Technical Changes

1. Added new value types: SENTRY_VALUE_TYPE_INT64 and SENTRY_VALUE_TYPE_UINT64 are added to the sentry_value_type_t enum.
2. Implemented new value creation functions: sentry_value_new_int64() and sentry_value_new_uint64() are introduced to create sentry_value_t instances holding 64-bit integer values.
3. Implemented accessor functions: sentry_value_as_int64() and sentry_value_as_uint64() are added to retrieve the 64-bit integer values from a sentry_value_t.
4. JSON serialization support: The sentry__jsonwriter_write_int64() and sentry__jsonwriter_write_uint64() functions are implemented to serialize the new value types into JSON format.
5. Msgpack serialization support: The value_to_msgpack function is updated to handle the new value types when serializing to msgpack format.
6. Internal representation: The thing_t struct's union is extended to accommodate int64_t and uint64_t values.
7. Type conversion: Implemented type conversion from int32 and double to int64 and uint64.

Architecture Decisions

The design maintains the existing tagged-pointer approach for sentry_value_t, extending the thing_t union to store the 64-bit integer values on the heap. This approach avoids increasing the size of the sentry_value_t struct itself, preserving ABI compatibility. The new value types are also frozen by default, similar to doubles and strings.

Dependencies and Interactions

These changes primarily affect the sentry_value API and its internal implementation. They also interact with the JSON and msgpack serialization components. No new external dependencies are introduced. The changes are designed to be backward-compatible with existing code that uses the sentry_value API.

Risk Considerations

1. Potential for data loss during type conversion: Converting from double to uint64_t involves truncation, which can lead to loss of precision. The code includes a TODO to address range validation, but this needs careful consideration.
2. Buffer overflow during string conversion: The snprintf calls in sentry__jsonwriter_write_int64 and sentry__jsonwriter_write_uint64 use a fixed-size buffer. While the size is likely sufficient, it's crucial to ensure that it can accommodate the largest possible 64-bit integer value to prevent buffer overflows. The buffer size in sentry__value_stringify should also be consistent with the JSON writer.
3. Double precision loss: Converting int64 and uint64 to double can result in precision loss due to the limited precision of double floating point numbers. This is handled by returning NaN if outside the double precision range, but this behavior needs to be clearly documented.

Notable Implementation Details

1. The THING_TYPE constants are defined using #define which can be error-prone. Consider using an enum.
2. The _int64_data and _uint64_data field names in the thing_t struct are not very descriptive. Consider renaming them to _int64 and _uint64 for clarity.
3. The tests include cross-type conversion tests, but more comprehensive boundary and edge case testing is needed, especially for double to uint64 conversion and for very large int64/uint64 values.

[supervacuus]

There are a couple of issues that require further deliberation. We can pair up on this if you'd like.

[jpnurmi]

Looks like sentry__value_from_json also needs some changes to avoid sentry__strtod_c and be able to handle > 2^53.