Drop fleet default min resources

[r4victor]

Closes #3585

Drop fleet default min resources both for run provisioning and instance provisioning paths. The main benefit is that users do not need to specify twice resources in run and fleet configurations unless they need to put extra constraints. Previously, to use instances with cpu, memory or disk lower than defaults, you had to lower the resources twice, and this could be hard to discover.

The new fleets behavior is that users get smallest available offers by default and specify resources if they need to lift/constraint the requirements. The run default min resources left unchanged to avoid provisioning runs on small offers by default.

Here's the new first fleet offers (without resources)

• aws: t3.nano, cpu=2 mem=0.5GB disk=100GB, $0.0052
• gcp: e2-highcpu-2, cpu=2 mem=2GB disk=100GB, $0.0495
• azure: Standard_F2s_v2, cpu=2 mem=4GB disk=100GB, $0.097
• nebius: cpu-e2 2vcpu-8gb, cpu=2 mem=8GB disk=100GB, $0.0496
• lambda: gpu_1x_a10, cpu=30 mem=214.8GB disk=1503.6GB gpu=A10:24GB:1, $1.29
• vultr: vc2-1c-1gb, cpu=1 mem=1GB disk=25GB, $0.007
• cudo: intel-broadwell, cpu=2 mem=3GB disk=100GB, $0.0286
• vastai: 29322263, cpu=8 mem=32GB disk=673.5GB gpu=RTX5070Ti:16GB:1, $0.267

For lambda and vultr disk size is tied to instance and cannot be changed so it's different than default 100GB fallback.

Existing fleets continue to have the defaults set. The change applies only to new fleets.