fix(techops-19980): resolve float32 precision loss in stream start_range and end_range (#39)

# Fix: Resolve float32 precision loss in stream start_range and
end_range

## Problem
The provider was experiencing precision loss when handling large
`start_range` and `end_range` values due to `float32` type conversion.
This caused Terraform to report "Provider produced inconsistent result
after apply" errors when values like `400990050` were converted to
`400990048` due to `float32`'s limited precision (24-bit mantissa).

## Root Cause
- OpenAPI specification defined `start_range` and `end_range` as
`*float32`
- `float32` can only precisely represent integers up to ~16,777,216
(2^24)
- Large block numbers (400M+) exceeded this precision limit
- Terraform detected the precision loss and flagged it as an
inconsistency

## Solution
Changed the API types from `*float32` to `*int64` for range fields:
- `StartRange *float32` → `StartRange *int64` 
- `EndRange *float32` → `EndRange *int64`
This eliminates precision loss while maintaining JSON compatibility
since JSON numbers can represent large integers

## Benefits
✅ **Eliminates precision loss** - `int64` can precisely represent all
blockchain block numbers

## Testing
- ✅ Provider compiled successfully
- ✅ Local testing with real QuickNode API
- ✅ Tested with problematic values: `400990050`, `400990100`,
`400990200`, `400990300`
- ✅ Verified `terraform plan` shows no changes after operations
- ✅ Confirmed create, update, and destroy operations work correctly

## Risks
⚠️ **OpenAPI specification mismatch** - If QuickNode API strictly
requires `float32`, this may cause issues (testing shows it works)
⚠️ **Code regeneration** - Changes will be overwritten if
`streams.gen.go` is regenerated from OpenAPI spec

Author: lwargin-circle