chore: bump github.com/oschwald/maxminddb-golang/v2 from 2.1.1 to 2.3.0

[dependabot]

Bumps github.com/oschwald/maxminddb-golang/v2 from 2.1.1 to 2.3.0.

Release notes

Sourced from github.com/oschwald/maxminddb-golang/v2's releases.

2.3.0

• This module now targets Go 1.25+.
• Reduced reflection decoding time and heap allocations on the hot path. A city-lookup benchmark decoding a geoip2-style result runs about 15% faster and allocates about 39% fewer bytes per lookup compared to 2.2.0.
• Specialized the IPv4 search-tree walk for 24-, 28-, and 32-bit record sizes to skip the IPv6 prefix when looking up IPv4 addresses.
• Decoding into a non-nil slice with sufficient capacity now reuses the caller's backing array instead of allocating a fresh slice, matching encoding/json semantics. Callers that share slice headers across Decode calls should be aware that the backing memory is now mutated.
• Reduced contention under concurrent lookups by switching the internal string cache to a lock-free design.

2.2.0

• Improved reflection decoding performance by skipping Unmarshaler checks for destination types that cannot implement the interface.
• Fixed verifier search-tree size arithmetic to match the reader's safe multiplication order instead of using an overflow-prone equivalent formula.
• Fixed unsigned bounds checks in search-tree node reads and traversal so very short malformed buffers return errors instead of underflowing the bounds calculation.
• Fixed the reflection decoder so pointer fields are not allocated when decoding fails with a type mismatch.
• Fixed Result.Prefix() to use the reader's measured IPv4 subtree depth instead of assuming IPv4 records always start at bit 96 in IPv6 databases.
• Fixed reflection decoding of negative int32 values into unsigned Go fields so it now returns a type error instead of wrapping them to large integers.
• Fixed lookups that followed malformed search-tree pointers past the data section so they now fail during Lookup instead of surfacing a deferred decode error.
• An error is returned when a maxminddb struct tag is clearly invalid (non UTF-8) instead of silently ignoring validation failures.
• Increased internal string cache size to 4096 entries to reduce cache thrashing and improve concurrent performance.

Changelog

Sourced from github.com/oschwald/maxminddb-golang/v2's changelog.

2.3.0 - 2026-05-17

• This module now targets Go 1.25+.
• Reduced reflection decoding time and heap allocations on the hot path. A city-lookup benchmark decoding a geoip2-style result runs about 15% faster and allocates about 39% fewer bytes per lookup compared to 2.2.0.
• Specialized the IPv4 search-tree walk for 24-, 28-, and 32-bit record sizes to skip the IPv6 prefix when looking up IPv4 addresses.
• Decoding into a non-nil slice with sufficient capacity now reuses the caller's backing array instead of allocating a fresh slice, matching encoding/json semantics. Callers that share slice headers across Decode calls should be aware that the backing memory is now mutated.
• Reduced contention under concurrent lookups by switching the internal string cache to a lock-free design.

2.2.0 - 2026-04-26

• Improved reflection decoding performance by skipping Unmarshaler checks for destination types that cannot implement the interface.
• Fixed verifier search-tree size arithmetic to match the reader's safe multiplication order instead of using an overflow-prone equivalent formula.
• Fixed unsigned bounds checks in search-tree node reads and traversal so very short malformed buffers return errors instead of underflowing the bounds calculation.
• Fixed the reflection decoder so pointer fields are not allocated when decoding fails with a type mismatch.
• Fixed Result.Prefix() to use the reader's measured IPv4 subtree depth instead of assuming IPv4 records always start at bit 96 in IPv6 databases.
• Fixed reflection decoding of negative int32 values into unsigned Go fields so it now returns a type error instead of wrapping them to large integers.
• Fixed lookups that followed malformed search-tree pointers past the data section so they now fail during Lookup instead of surfacing a deferred decode error.
• An error is returned when a maxminddb struct tag is clearly invalid (non UTF-8) instead of silently ignoring validation failures.
• Increased internal string cache size to 4096 entries to reduce cache thrashing and improve concurrent performance.

Commits

• a48a9be Merge pull request #210 from oschwald/dependabot/go_modules/golang.org/x/sys-...
• bf13761 Bump golang.org/x/sys from 0.41.0 to 0.44.0
• 4b96bed Merge pull request #211 from oschwald/greg/perf-improvements
• a37ef6b Scope goconst suppression to tests
• 79cd080 Run modernize on the code
• def3ca5 Target Go 1.25
• 52a7da6 Add 2.3.0 changelog entry
• 496ae65 Add reflect.Uint fast path to typed decoder
• 6eaf764 Precompute interface-field unmarshaler dispatch
• b8a7100 Specialize IPv4 search-tree walk
• Additional commits viewable in compare view