Change poet client to cache the full response of the v1/info

[fasmat]

Description

From a discussion on PR #6116: #6116 (comment)

The whole info response should be cached instead of just the types.CertifierInfo field of the response.

Affected code

The poet client already caches part of the response returned from the v1/info endpoint via these two methods:

go-spacemesh/activation/poet.go

Lines 618 to 641 in 15ebc7b

	func (c *poetService) Certify(ctx context.Context, id types.NodeID) (*certifier.PoetCert, error) {
	if c.certifier == nil {
	return nil, errors.New("certifier not configured")
	}
	info, err := c.getCertifierInfo(ctx)
	if err != nil {
	return nil, err
	}
	return c.certifier.Certificate(ctx, id, info.Url, info.Pubkey)
	}
	func (c *poetService) getCertifierInfo(ctx context.Context) (*types.CertifierInfo, error) {
	info, err := c.certifierInfoCache.get(func() (*types.CertifierInfo, error) {
	certifierInfo, err := c.client.CertifierInfo(ctx)
	if err != nil {
	return nil, fmt.Errorf("getting certifier info: %w", err)
	}
	return certifierInfo, nil
	})
	if err != nil {
	return nil, err
	}
	return info, nil
	}

This should be changed so that the whole response is cached:

func (c *poetService) getInfo(ctx context.Context) (*types.PoetInfo, error) {
	info, err := c.infoCache.get(func() (*types.PoetInfo, error) {
		info, err := c.client.Info(ctx)
		if err != nil {
			return nil, fmt.Errorf("getting info: %w", err)
		}
		return info, nil
	})
	if err != nil {
		return nil, err
	}
	return info, nil
}

simplifying verifyPhaseShiftConfiguration to the following:

func (c *poetService) verifyPhaseShiftConfiguration(ctx context.Context) error {
	info, err := c.getInfo(ctx)
	if err != nil {
		return err
	}

	if info.PhaseShift != c.expectedPhaseShift {
		return ErrIncompatiblePhaseShift
	}

	return nil
}

[acud]

A note for consideration is etags. They're meant to allow for resource expiration at-the-server and allow intermediate layers to do caching and invalidation more easily. Just caching for some time might backfire under certain edge cases (please do consider my ignorance over the topic because I'm not sure they exist here).

[fasmat]

Sure, but this is just to keep track of a change that we excluded from a previous PR to get it merged quicker 🙂. PoET doesn't send Etags, the primary reason we are caching at all is to reduce the number of requests from the node to the servers. The node sends the same request for each identity to the same PoET (usually) around the same time, this just ensures that in that case we re-use the response we already got for the first identity for all others.

[poszu]

Implemented in #6274